Coronavirus (COVID-19) #LivUniCOVID

Signals and Communication Technology
Vishram Mishra
Jimson Mathew
Chiew-Tong Lau
QoS and Energy

Management
in Cognitive
Radio Network
Case Study Approach
More information about this series at http://www.springer.com/series/4748
Vishram Mishra Jimson Mathew
•
Chiew-Tong Lau
QoS and Energy

Management in Cognitive
Radio Network
Case Study Approach
123
Vishram Mishra Chiew-Tong Lau
Nanyang Technological University Nanyang Technological University
Singapore Singapore
Singapore Singapore
Jimson Mathew
Indian Institute of Technology Patna
Patna
India
ISSN 1860-4862 ISSN 1860-4870 (electronic)

ISBN 978-3-319-45858-8 ISBN 978-3-319-45860-1 (eBook)
DOI 10.1007/978-3-319-45860-1
Library of Congress Control Number: 2016952004
© Springer International Publishing Switzerland 2017

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or
dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt
from the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained
herein or for any errors or omissions that may have been made.
Printed on acid-free paper
This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
The approval of free usage of frequency spectrum via dynamic spectrum access
methods on a non-interfering policy basis in various countries has made the cog-
nitive radio technology a prominent candidate for wireless communication systems.
Most of the wireless devices and their supporting infrastructure available today are
capable of handling multimedia content delivery over wireless channels. Therefore,
quality of service (QoS) is an important parameter for users experience while
utilizing the services rendered on their wireless devices. For realization of cognitive
radio-based dynamic spectrum access over conventional devices with conventional
services (VoIP, Video, text messaging, Web page, etc.), QoS is an important factor
that need to be provided to the wireless devices and applications. The QoS is
broadly defined via two network parameters: throughput and latency. Similarly,
energy management is also an important criteria for increasing the lifetime of
portable wireless devices. Different aspects of a wireless network contribute toward
varying network parameters and energy consumption, with each aspect having its
own weight in contribution.
In this book, we discuss a QoS provisioning framework for cognitive radio
network, which identifies and divides attributes of system into different modules
responsible for quality of service and energy management. These modules are
independently studied and a relation between them is shown to depict the inter-
action between these modules. These modules are responsible for channel selection,
media access control, self-co-existence, and energy saving. A relationship is drawn
among these modules for a coherent decision followed by the system. The
framework supports four access categories for service differentiation. The frame-
work is also designed in such a way that it separates main communication protocol
stack from quality of service engine and thus allows its support for legacy protocol
stack with minimal variation. Along with the provisioning of QoS, energy man-
agement in cognitive radio is also discussed. Since, the goal of the cognitive radio is
to enter into a commercial market in a transparent way, energy management of the
devices in the secondary network plays a crucial role to achieve this. Energy
management from the viewpoint of channel selection, spectrum sensing order, and
media access scheme is also discussed in this book. Each of these viewpoints
v
vi Preface
provides different ways through which energy can be efficiently managed in

cognitive radio network.
Chapter 4
Spectrum Management
in CRN Framework
Chapter 5
Chapter 3 Media Access in CRN

Chapter 2 Framework
Chapter 1 QoS and Energy
Cognitive Radio Management
Introduction Network - a Review Framework for
CRN Chapter 6
Energy Management
in CRN Framework
Chapter 7
Self Coexistence in
CRN Framework
The structure of this book is given as shown in figure above. It is recommended

for a beginner to start from Chap. 1, while an experienced reader in the field of
cognitive radio can directly proceed from Chap. 3.
Chapter 1 introduces basic motivation for reading this book and provides a basic
background on cognitive radio and spectrum regulation issues. This chapter
reminds the need of opportunistic spectrum usage and presents a comparison of
software-defined radio with cognitive radio. This chapter also introduces basics on
QoS, channel selection techniques, MAC protocols, and self-co-existence in cog-
nitive radio network. Chapter 2 provides a review on different modules of the
cognitive radio framework, including spectrum management, spectrum selection,
media access schemes, and cognitive radio platforms. These modules are discussed
from the perspective of quality of service and energy management.
Chapter 3 briefly describes QoS provisioning and energy management frame-
work for cognitive radio network. QoS provisioning framework for cognitive radio
network at Layer 2 of the protocol stack and different functional modules for QoS
provisioning from the perspective of latency and throughput are discussed in this
chapter. It is shown that latency and throughput can be directly identified by
the media access scheme and channel selection technique, respectively, and also the
effect of self-co-existence and energy management is separately studied.
A case study is conducted on spectrum usage behavior in Chap. 4. The concept
of deterministic and stochastic-based classification of channel selection technique is
discussed. Deterministic channel selection schemes utilize deterministic behavior of
spectrum and make decision on spectrum selection ahead of time.
A packet-to-channel tabular mapping-based fast spectrum selection scheme is
discussed for stochastic channel environment.
In Chap. 5, a case study is done on a hybrid media access control protocol for
QoS provisioning which encompasses two different channel access etiquettes. The
Preface vii
discussed media access scheme realizes service differentiation for four different
types of packets. The support for important network operations such as broadcast
and multicast, as well as realization of power saving mode is also discussed.
A case study is presented on the energy management functionality of cognitive
radio network via utilizing a spectrum selection strategy in Chap. 6. A scheme is
discussed to allow selection of low bandwidth channels so as to conserve trans-
mission energy. The scheme also realizes a packet admission control technique for
graceful degradation of QoS of the system.
Resource relocation and resource sharing-based self-co-existence techniques are
discussed in Chap. 7. The approach of graph coloring is utilized to allocate disjoint
set of resources, so as to mitigate interference. A frame sharing based approach is
also discussed for self-coexistence.
Singapore, Singapore Vishram Mishra

Patna, India Jimson Mathew
Singapore, Singapore Chiew-Tong Lau
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Why This Book? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Spectrum Regulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.1 Licensed Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.2 Unlicensed Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.3 Open Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Opportunistic Spectrum Usage . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Software Defined Radio and Cognitive Radio . . . . . . . . . . . . . . . 9
1.4.1 IEEE Groups Working on Spectrum Sharing . . . . . . . . . 10
1.4.2 Cognition Cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4.3 Cognitive Engine and Framework . . . . . . . . . . . . . . . . . . 12
1.4.4 Cognitive Radio Network . . . . . . . . . . . . . . . . . . . . . . . . 15
1.5 Quality of Service (QoS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.5.1 QoS Provisioning for Latency Guarantee . . . . . . . . . . . . 18
1.5.2 QoS Provisioning for Throughput Guarantee . . . . . . . . . . 20
1.6 Channel Selection Techniques in Cognitive Radio Network . . . . . 20
1.6.1 Channel Selection in CR Based Infrastructure
Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.6.2 Channel Selection in CR Based Ad-hoc Network . . . . . . 22
1.7 MAC Protocols for Cognitive Radio Networks. . . . . . . . . . . . . . . 23
1.7.1 Random Access Based MAC Scheme . . . . . . . . . . . . . . . 24
1.7.2 Time-Slotted Based MAC Scheme . . . . . . . . . . . . . . . . . 26
1.8 Self-coexistence in Cognitive Radio Networks . . . . . . . . . . . . . . . 28
1.8.1 Resource Relocation Based Self-coexistence . . . . . . . . . . 29
1.8.2 Resource Sharing Based Self-coexistence . . . . . . . . . . . . 30
1.9 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
ix
x Contents
2 Cognitive Radio Network- A Review . . . . . . . . . . . . . . . . . . . . . . .... 39

2.1 Spectrum Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .... 39
2.1.1 Ant Colony Optimization Based Spectrum
Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .... 39
2.1.2 Non-linear Optimization Based Spectrum
Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .... 41
2.1.3 Game Theory Based Spectrum Management . . . . . . .... 46
2.1.4 Learning Automata Based Spectrum Selection . . . . .... 51
2.1.5 Spectrum Selection in Varying Channel Bandwidth
Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.2 Media Access Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.2.1 QoS Aware Media Access Schemes . . . . . . . . . . . . . . . . 56
2.2.2 High Throughput Media Access Schemes . . . . . . . . . . . . 64
2.2.3 Self-coexistence Based MAC Protocol . . . . . . . . . . . . . . 67
2.3 Energy Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.3.1 Cooperative Sensing Based Energy Efficient
Spectrum Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . .... 72
2.3.2 Non-cooperative Sensing Based Energy Efficient
Spectrum Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
2.4 Cognitive Radio Platforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
2.4.1 From FPGAs to Software Defined Radio . . . . . . . . . . . . 89
2.4.2 From Software Defined Radio to Cognitive Radio . . . . . 89
2.4.3 SDR Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
2.4.4 WARPnet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
2.4.5 RTL-SDR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
2.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3 QoS Provisioning and Energy Management Framework
for CRN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
3.1 QoS Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3.1.1 Latency Versus Throughput. . . . . . . . . . . . . . . . . . . . . . . 98
3.1.2 Self-coexistence and Its Role in QoS . . . . . . . . . . . . . . . 99
3.1.3 Energy Management and QoS . . . . . . . . . . . . . . . . . . . . . 99
3.2 QoS Framework for Cognitive Radio Network . . . . . . . . . . . . . . . 100
3.3 Detailed Layer 2 QoS Provisioning Framework . . . . . . . . . . . . . . 101
3.3.1 Mode of Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
3.3.2 Generic Protocol Stack . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.4 Self Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
3.4.1 Channel Availability Model . . . . . . . . . . . . . . . . . . . . . . . 108
3.4.2 Quiet Periods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Contents xi
4 Case Study: Spectrum Management in CRN Framework . . . . . . . . . 111

4.1 Spectrum Usage Behavior. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.1.1 Deterministic Usage Behavior . . . . . . . . . . . . . . . . . . . . . 111
4.1.2 Stochastic Usage Behavior . . . . . . . . . . . . . . . . . . . . . . . 112
4.2 Reconfigurable Channel Selection . . . . . . . . . . . . . . . . . . . . . . . . . 113
4.3 Channel Selection in Deterministic Environment . . . . . . . . . . . . . 114
4.3.1 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
4.3.2 Deterministic Learning with Spectrum Selection
and Usage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .... 116
4.3.3 Minimal Channel Switch Requirement . . . . . . . . . . .... 118
4.3.4 Maximum Throughput Requirement . . . . . . . . . . . . .... 120
4.3.5 Intermediate Solution to Provide High Throughput
Along with Minimal Channel Switch Requirement . . . . . 120
4.3.6 Complexity Analysis of Algorithms . . . . . . . . . . . . . . . . 121
4.3.7 Spectrum Usage. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
4.3.8 Performance Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4.3.9 Ad-hoc Mode Operation . . . . . . . . . . . . . . . . . . . . . . . . . 127
4.4 Channel Selection in Stochastic Environment . . . . . . . . . . . . . . . . 128
4.4.1 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
4.4.2 Communication Segment . . . . . . . . . . . . . . . . . . . . . . . . . 129
4.4.3 Spectrum Decision and Mapping of Packets . . . . . . . . . . 129
4.4.4 Load Balancing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
4.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
5 Case Study: Media Access in CRN Framework . . . . . . . . . . . . . . . . . 139
5.1 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
5.1.1 Traffic Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
5.1.2 Channel Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
5.1.3 Quiet Period Distribution . . . . . . . . . . . . . . . . . . . . . . . . . 141
5.2 Hybrid Media Access Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
5.2.1 Arbitration Interframe Spaces . . . . . . . . . . . . . . . . . . . . . 143
5.3 Initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
5.4 Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
5.4.1 Data Transfer on Reservation Based Channels . . . . . . . . 145
5.4.2 Data Transfer on Contention Based Channels . . . . . . . . . 145
5.4.3 Data Transfer on a Foreign Channel . . . . . . . . . . . . . . . . 146
5.5 Power Saving Mode Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
5.6 Broadcast and Multicast Operation . . . . . . . . . . . . . . . . . . . . . . . . 147
5.7 Performance Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
5.8 Infrastructure Mode Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
5.9 Hidden Terminal Problem in Ad-hoc Mode . . . . . . . . . . . . . . . . . 152
5.10 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
xii Contents
6 Case Study: Energy Management in CRN Framework . . . . . . . . . . . 155

6.1 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
6.2 Energy Aware Spectrum Allocation Scheme . . . . . . . . . . . . . . . . 157
6.2.1 Spectrum Decision. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
6.2.2 Bare Bandwidth Calculation . . . . . . . . . . . . . . . . . . . . . . 158
6.2.3 Admission Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
6.3 Performance Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
6.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
7 Case Study: Self-coexistence in CRN Framework . . . . . . . . . . . . . . . . 167
7.1 Self-coexistence Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
7.1.1 Detection of Interfering Network . . . . . . . . . . . . . . . . . . . 168
7.1.2 Recovery Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
7.2 Resource Relocation Based Self-coexistence. . . . . . . . . . . . . . . . . 169
7.2.1 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
7.2.2 Graph Coloring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
7.2.3 Optimization Problem Equivalent of ERMC . . . . . . . . . . 172
7.2.4 Relationship Between Optimization Problems . . . . . . . . . 173
7.2.5 Self Coexistence Scheme with QoS Provisioning . . . . . . 175
7.2.7 Infrastructure Mode Operation . . . . . . . . . . . . . . . . . . . . . 179
7.3 Resource Sharing Based Self-coexistence . . . . . . . . . . . . . . . . . . . 180
7.3.1 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
7.3.2 Design and Working . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
7.3.3 Multiple CR Network Coexistence . . . . . . . . . . . . . . . . . 185
7.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Appendix A: Proof of Lemma 4.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
Appendix B: Proof of Lemma 4.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Appendix C: Proof of Lemma 4.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Appendix D: Proof of Polynomial Bound Convergence
of Algorithm 7.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
Chapter 1
Introduction
Recently, the rapid proliferation of portable wireless communication devices sup-

porting multimedia content playback and online gaming have emerged in market.
The applications supported by these devices consumes hefty bandwidth for which
a wider spectrum band is utilized. An example of historical increase in requirement
of wider band can be seen in development of market from 2G to 5G generation of
mobile networks. The most useful spectrum bands are in the range of sub 3-GHz;
which already have been allocated for specific purposes and are regulated by the
regulatory authority of the respective country. The conventional method of channel
allocation follows fixed channel allocation (FCA) policy; which implies allocation
of a particular band of spectrum, to an operator/user for a given purpose [1]. The
static allocation of channels by FCA policy has resulted in exhaustion of available
usable spectrum as the number of bandwidth hungry applications keeps in increas-
ing while there is no room left for available spectrum in the usable band because of
overcrowding. This has also resulted in technological advancements to utilize higher
frequency bands (60 GHz range); but these suffer from problems of high attenuation
and line-of-sight operation.
These large number of portable wireless communication devices utilize spectrum
over non-uniform space and time, leading to a significant amount of unused frequency
spectrum. Studies sponsored by various spectrum regulators in their respective coun-
tries have shown that the more than 70 % of the spectrum allocated to licensed user
is not utilized [2]. Therefore there exist a paradox between the underutilization of
the spectrum and its scarcity.
The underutilization of spectrum results in spectral holes, also called as white
spaces. These spectral holes can be opportunistically exploited by an opportunistic
(unlicensed) user, which ultimately leads to a solution for spectrum scarcity. An
opportunistic user hops from one spectral hole to another while mitigating inter-
ference with incumbent (licensed) users (as shown in Fig. 1.1) [3]. To realize the
functionality of an opportunistic user, cognitive radio (CR) based dynamic spectrum
access technique has been formulated. The basic idea is to employ a hierarchical
© Springer International Publishing Switzerland 2017 1
V. Mishra et al., QoS and Energy Management in Cognitive Radio Network,
Signals and Communication Technology, DOI 10.1007/978-3-319-45860-1_1
2 1 Introduction
Spectral holes
Time
Fig. 1.1 Opportunistic utilization of spectral holes
model, where licensed (primary/incumbents) and unlicensed (secondary/CR) users

coexist in the same frequency band such that the primary users are oblivious of
the presence of secondary users [4]. A secondary user first senses the channel, and
if it is unoccupied then only it performs transmission of data, thereby maintaining
interference with primary users to a minimum level.
To accomplish the task of opportunistic utilization, a cognitive radio has to go
through the process of Spectrum Sensing, Spectrum Decision and Spectrum Shar-
ing [5]. Spectral holes in the primary bands are detected with the help of spectrum
sensing functionality of a cognitive radio. Since results from spectrum sensing are
critical in mitigating interference with licensed users, it is of paramount importance
to obtain precise information about spectral opportunities. This requires cooperation
among different secondary users to avoid false alarm and missed detection [6, 7].
After the detection of spectral holes among different primary heterogeneous bands,
a decision is made on allocation of primary channels based on different constraints
imposed by the user. Primary bands are governed by local regulatory agencies and
hence local policies enforced on them varies [8]. Policies like maximum accessi-
ble bandwidth, transmit power and transmit duration are main determining factors
for channel selection [9]. The channels are allocated in such a way that the quality
of service (QoS) requirements of different applications are met. Different channel
allocation strategies try to cater to different needs of users. The performance and
competency of a channel selection technique depends upon various factors includ-
ing number of secondary users, primary user interference, number of channels used
and algorithm used [10]. When different users exist on the same channel, an access
scheme is required to provide a coordination between different users so that access
collisions can be alleviated. Several approaches have been proposed to develop effi-
cient media access control (MAC) schemes for cognitive radio based network [11].
But most of these MAC protocols solve a particular problem and do not provide a
generalized framework for cognitive radio system.
Since there is an out-burst of multimedia and voice over IP (VoIP) traffic, an
important objective during the operation of network is to ensure a minimum grade
of service to such delay sensitive traffic [12]. In a traditional network, all users are
designated with equal priority to access a common resource (bandwidth). Therefore
1 Introduction 3
if the total demand of resource by traffic from all users in the network exceeds the
available resource, then the throughputs of all users are equally impeded. This has
varying impact on user experience depending upon the delay-sensitivity of the users
application. For example, one second delay is acceptable in web browsing while the
same is not acceptable for VoIP users. To mitigate the adverse effects of resource
equalization, traffic from different users are prioritized and appropriately handled.
Some of the anticipated applications that require special handling of the traffic are
VoIP, video streaming, music streaming and interactive gaming [13]. The special
handling of traffic is evaluated using quality of service (QoS), which is defined as
the ability of a system to provide a guarantee on assured grade of service [14]. A
different context to service quality is Quality of Experience (QoE) which is used to
describe the perception of the consumer about usefulness of a system [15]. Currently,
the development in the field of cognitive radio solves localized problem in their
respective domains without any inter-relation. A framework is required which can
integrate different modules responsible for QoS provisioning in the cognitive radio
system. In this book, case studies are done about a QoS provisioning framework
for cognitive radio network which works at the Layer 2 of the protocol stack while
integrating the functionality of spectrum decision, medium access control and self-
coexistence.1
1.1 Why This Book?
The fundamental goal of this book is to discuss a QoS provisioning framework for
cognitive radio network which provides QoS support at Layer 2. For the purpose
of same, different technologies responsible for QoS have been identified in cogni-
tive radio network and a relationship is established among them. The framework
comprises of multiple units, each of which is individually responsible for QoS pro-
visioning in the system as a whole. The coordination among these units (blocks) is
discussed to enable the system to perform its intended task. These units are connected
in a specified manner, with each connection either uni-directional or bi-directional
to indicate the type of relation among different units. The configuration, queries or
results are passed from one unit to other unit. The configuration passed to a block
contains the parameters to configure it so as to achieve the desired utility from it. The
query consists of commands asking for data or result. An example of query would
be channel state and its response would be either free or busy. The query-response
blocks have bi-directional link between them. The framework works in parallel with
the main communication protocol stack and therefore does not directly alter the
behavior of system. This results in minimum dependency with easy integration of
framework in a conventional communication system protocol stack. The compart-
1 Layers of the protocol stack refers to the architectural layers of the Open System Interconnection
model proposed by the International Organization for Standardization for the purpose of standard-
ization of the internal functions of a communication system [16].
4 1 Introduction
mentalization of different units allows easy abstraction, debugging and maintenance

of the system as a whole.
The framework works by providing substitution of protocol stacks in conventional
communication system with its own instances. An example would be generation of
MAC protocol instance on the fly depending upon the requirement. This will help to
classify different activities involved in a QoS framework while dividing the overall
work of the QoS framework into smaller units. These units are independently studied,
improved and created if they do not exist elsewhere.
The objectives of this book are as follows:
1. Classification and compartmentalization of different modules that are responsible
for QoS provisioning at Layer 2 of the protocol stack. The identification of these
modules depends upon the type of QoS required: latency guarantee or throughput
guarantee. These guarantees are handled by their respective modules and hence
compartmentalization allows implementation of abstract behavior of different
functionalities provided by the system.
2. Since QoS provisioning modules at Layer 2 are compartmentalized, a relationship
need to be defined for the purpose of requirement translation. The guarantee on
throughput and latency is translated over these relationship.
3. Each of these modules is independently studied, researched and new techniques
are developed to provide better performance than that of its previous predecessor.
These modules are incorporated with new interfaces so as to allow interaction
with other modules of the Layer 2 of the protocol stack.
4. The performance of each module is independently evaluated which allows clear
performance analysis. And it allows developers to give an idea about the maximum
performance achievable by the module; so a translation of performance can be
done while integrating the module into a system.
1.2 Spectrum Regulation
International Telecommunication Union (ITU) [17] is a specialized agency of the

United Nations (UN) responsible for the coordination of shared global radio spectrum
in the range of 3 KHz–300 GHz as well as the evolution and coordination of different
wireless standards. According to ITU each country/state has its sovereign right to
regulate its telecommunication. Therefore each country may have its own set of
standards laid down besides the standards defined by ITU.
Spectrum regulation may vary from country to country, but they all can be clas-
sified into three broad categories:
1. Licensed Spectrum: This type of spectrum is licensed to a particular operator
and enforced through a regulator [2]. This can be further classified into exclusive
usage and shared usage. Exclusive usage of licensed spectrum allows exclusive
usage rights to the licensee, whereas shared usage of licensed spectrum allows
two or more operators to utilize licensed channel on a coordinated basis.
1.2 Spectrum Regulation 5
2. Unlicensed Spectrum: This type of spectrum is available for free to all users, who
conform to the standards defined for accessing the particular band of spectrum.
Unlicensed National Information Infrastructure (U-NII) radio band is an exam-
ple of unlicensed spectrum [18]. Industrial, Scientific and Medical (ISM) radio
band, whose applications including RF heating, microwave heating and medical
diathermy machines, is also part of unlicensed spectrum. Recently ISM bands
have been widely utilized by non-ISM applications including wireless sensor
network, wireless LANs, cordless, Radio Frequency Identification (RFID) and
Short Range Devices (SRD) on various operating frequencies, all for the purpose
of communication.
3. Open Spectrum: Open spectrum refers to the methodology of accessing any
band of spectrum without any permission under the constrained set of standards,
protocols and maximum allowed interference [19, 20]. The overall objective of
this technology is to effectively utilize spectrum on spatial basis without causing
any harmful interference to operating incumbents.
1.2.1 Licensed Spectrum
Licensed channels are allocated via Fixed Channel Allocation (FCA) policy, which
implies allocation of a particular block of spectrum to an operator/user for a given
purpose [1]. Allocation of channels for exclusive usage results in inefficient spectrum
utilization, because even if the spectrum is underutilized or unused, it cannot be
accessed by other users. An example of FCA policy is the allocation of 2G and 3G
spectrum for cellular communication in various countries. The spectrum allocation
chart of Singapore is shown in Fig. 1.2.2
Since most part of the spectrum has already been allocated to existing com-
munication technologies, new technologies demand for more bandwidth. Although
technological advancements are pushing towards higher frequency bands (in the
range of ∼60 GHz), the operation at these bands suffers from high attenuation and
requirement of line-of-sight operation. The high attenuation results in requirement
of high transmission power making end-user energy inefficient. Similarly, require-
ment of line-of-sight operation restricts the design and operation of the end-user in a
network. To overcome the problem of licensed FCA policy, spectrum trading tech-
niques have been proposed as a secondary market for spectrum ownership and usage
[21]. In spectrum trading, spectrum can be leased temporarily, sold permanently or
traded as Future and Options [22]; aiming to increase spectrum utilization.
1.2.2 Unlicensed Spectrum
Some part of spectrum is allocated by the spectrum regulators for the purpose of unli-
censed operation so as to allow users to access spectrum on non-licensee basis. Even
2 Singapore Allocation Chart, Infocomm Media Development Authority. Retrieved from www.imda.
gov.sg
3 MHz
3 kHz -
30 GHz
30 MHz
– 3 GHz
3 GHz –
300 kHz
3 MHz –
300 MHz
300 GHz
300 MHz
30 MHz –
30 GHz –
30GHz 3GHz 300MHz 30MHz 3MHz 300kHz 3kHz

Rad Nav
31 -MET 3.1 - 328.6 - 37.5 - 3.155 - Aeronatical
RadioLocate Aero Radio 3.4 -
Fixed 3.4 - 47 - 3.5 - Aeronatical
3.9 - Fixed
Fix Sat 335.4 -
31.3 - 3.95 - Aeronatical
ISM Fix
4.2 -
Land Mobile
Low Power
4- Maritime
Maritime
Aero Rad 74.6 - 4.438 -

/ Satellite
LM
4.4 - 74.8 -
Fixed
4.65 -
4.5 - Aero Rad 4.75 -
75.2 - 4.8 - Fixed
Fix Sat 399.9 -
Radionavigation
Radionavigation
Fixed
4.8 - 400.05 - Radio Sat 75.4 - 4.85 -

4.99 - 400.15 - 4.995 - Mobile
5 - Radio Astro 5.005 -
LM
9-
Met
300 kHz - Aeronautical Radionav
5.25 - Aero Rad 406.1 - Fixed

Fixed
RadioLocate 410 - Radio Astro 5.45 - 415 -

5.35 -
Land Mobile
Aeronatical
LM
Aero Rad 5.73 -

Fixed
36 - 88 -
Fixed
5.47 - 5.95 - Fixed

M Radio 430 -
5.65 - 6.2 - Broadcast
LP
FM
Aeronatical
434.6 -
Fixed Maritime
108 - 6.525 -
Rad
5.725 -
Radionavigation
5.85 - 6.765 - Aeronatical

ISM
ISM
40.5 - 7-
Mobile
Fixed
5.875 - 7.1 - Amateur

Broadcast
Maritime Mobile
470 - 118 - 7.35 -

LP
Fixed
LM
494 - 8.195 - 526.5 -

Mobile
Maritime
Radionav Mobile
8.815 -
Fixed
Fix Sat
Aeronatical
Aero Aeronatical
Radionavigation
137 - 9.04 - Aeronatical

Met Mob Fixed
Mobile / Satellite
138 - 9.4 -
9.9 - Broadcast
7.075 - 9.995 - Fixed
10.005 -
Broadcast
Fixed
Fixed
59 - 10.1 - Aeronatical
7.25 - 10.15 -
Land Mobile
Fixed
Fix Sat 144 - 11.175 -
ISM
7.75 - Amateur 11.4 - Aeronatical

64 -
Broadcast
7.9 - 790 - 146 - Fixed 130 -

11.6 -
Digital Audio
Fixed Satellite
FS
Broadcast
Fixed
Mobile
Fixed
8.4 - 890 - 12.1 -

Fixed
12.23 - Fixed
Low Power
8.75 -
Land Mobile
Maritime
149.9 - 13.2 -
Mobile
960 -
Medium Wave Broadcasting
8.85 - 150.05 - Radio Sat 13.36 - Aeronatical

RL
LM Fix 13.57 -
Rad AR MR
Broadcast
Rad
Aero
9- 156 -
Aero Rad 1215 - 13.87 -
9.2 - Maritime Fixed
Maritime
MR
9.3 - RadioLocate
157.45 - 14 -
EGSM 160.625 - LM Fix 14.35 - Amateur 1606.5 -
116 - 1300 - Maritime Fixed
Radionavigation
Mobile / Satellite
Aeronatical
Maritime Mobile
162 - 14.99 - Mobile

Radionav
1350 -
Radiolocate
10 - LM Fix 15.01 - 1800 -

RadioLocate
10.5 - Ama 1400 - 174 - 15.1 - Aeronatical
Fixed Broadcast
ISM
10.68 - 1429 - 15.8 -
10.7 - 1452 - Fixed 160 -
11.7 - FS 16.36 -
144 - BS Broadcast
12.2 - 1492 - Maritime
Amateur
12.5 - 1525 - 17.41 -
Amateur
Mob Sat 17.48 - Fixed

Radionav
1530 - Broadcast
Fixed
Distribution Systems
17.9 -
Multipoint Microwave
Broadcasting / Satellite
Fixed Sat MMS Aeronatical
1544 - 18.03 - 2000 -
DAB
13.25 - Mob Sat Fixed

13.4 - Aero Rad 1545 - 18.068 -
Low Power
AMS
Broadcasting
1555 - 18.168 - Amateur

13.75 - 1559 - LMS Fixed
Aero 18.78 -
GPS 2065 -
Rad Maritime
Fixed
1610 - 18.9 -
Broadcasting
Medium Wave
Aero
Mobile Satellite
Broadcast
Fixed Sat
MS 19.02 -
Rad
1628.5 - Fixed
Mobile Mobile
14.8 - 230 - 19.68 -

Land Maritime
15.35 - Fixed 1631.5 - Maritime 2107 -

15.4 - MMS 19.8 -
15.7 - Aero Rad 1645.5 - MS 19.99 - Fixed
Fixed
20.01 -
1646.5 -
AMS Fixed
Aero Radionav
Mobile
Fixed
1656.5 - 235 - 21 -
Radiolocate
17.3 - LMS 21.45 - Amateur 2170 -
Land Mobile
1660 - Broadcast
Radiolocation
17.7 - 1670 - 21.85 -
Met Mob Fixed
1675 - 21.924 -
Mobile
Met Sat
Maritime
1710 - 22 - Aeronatical
Maritime 2194 -
Fixed
22.855 -
Mob
Fixed
1980 -
Fixed Satellite
Mob Sat
Fixed
23.2 -
Aeronautical Radionav
2010 -
Aeronatical
Mobile
/ Satellite
21.2 -
Mob
241 - 23.35 -
Fixed
2110 - 2495 -
21.4 - Mobile Fixed
24.89 - 2505 -
2170 -
Radio Astronomy
Amateur
Metrological Aids
Mob Sat 24.99 -
2200 - 25.01 -
ISM
25.07 - Fixed
Fixed
Fixed
2400 - Maritime
25.21 -
Fixed
Low Power
275 -
Mobile
Land Mobile
Mob ISM
Singapore Spectrum Allocation Chart
24 - 2483.5 - Fixed
ISM
ISM Ama Mob Sat 25.55 -
24.25 - 2500 - Radio Astro 2850 -
Fix MDS 25.67 -
Low Power
2535 - Broadcast
Fix BS 26.1 - 285 -
27.5 - 2655 - Maritime
Fixed 26.175 -
2690 - Fix ISM
Fig. 1.2 Spectrum allocation chart of Singapore. Almost all frequency bands are allocated up to 30 GHz (Adapted from www.imda.gov.sg)
2700 - 27.5 -
28.5 -
Fixed
Aero Radio 28 -
•Frequency Spectrum
is not drawn to scale.
GPS
2900 -
Maritime
29.7 - Amateur
Radionav
Fix Sat
29.5 - Radionav Fix
Aeronatical Mobile
300GHz 30GHz 3GHz 300MHz 30MHz 3MHz 300kHz
/ Prohibited
To be Planned
1 Introduction 6
1.2 Spectrum Regulation 7
if this type of spectrum is free-to-use, it is regulated in terms of interference policies

which are in turn governed by limitation on transmission power and conforming to
the standards of coexistence capabilities. The type of telecommunication users oper-
ating on such frequency bands are generally termed as Short Range Devices (SRD).
SRD generally operate with transmission power less than 100 mW of Equivalent
Isotropic Radiated Power (EIRP). An example of unlicensed spectrum in Singapore
is the allocation of 2.4000–2.4835 GHz band for Wireless Video transmitter, Blue-
tooth, WLAN and other SRDs [23]. All devices operating in this band are limited
to 100 mW EIRP of transmission power, except for WLAN operation where the
transmission power is limited to 200 mW EIRP [23].
1.2.3 Open Spectrum
The success of unlicensed bands has resulted in plethora of devices to operate in this
band, resulting in overcrowding and hence leading to a problem termed as “tragedy
of commons” [24]. In other words, over-utilization of a particular band of spectrum
implies high interference thus making it less effective in terms of operation. To
mitigate such a scenario, open spectrum techniques are proposed. Open spectrum
allows users to access any portion of spectrum under the given constrained set of
etiquettes, standards and maximum allowed interference, that need to be enforced for
the purpose of spectrum sharing [19]. This allows users to access spectrum without
going through the process of licensed spectrum allocation by regulatory agencies.
Two major form of open spectrum access techniques have been discussed in the
literature:
1. Underlay Spectrum Access: This type of spectrum access refers to the simulta-
neous secondary uncoordinated utilization of spectrum even in the presence of
licensed incumbents on the spectrum. The main idea behind this technique is to
keep the transmission power low, so that the interference received by the licensed
incumbents is below a certain threshold [25]. To achieve an effective communica-
tion, the transmitted signal is spread over a wide band of spectrum with low power
transmission. Various spread spectrum techniques can be applied to achieve this
including Ultra-Wide Band (UWB) which has become popular for its utilization
in system such as Wireless USB (WUSB) and Wireless Firewire [26]. This type
of spectrum access do not require development of new protocols.
2. Overlay Spectrum Access: This type of spectrum access refers to the utilization
of under-utilized spectrum by a secondary user while mitigating interference with
licensed incumbents; a technique referred as opportunistic utilization [27, 28]. To
realize the implementation of such secondary user, development of new protocols
is required. Cognition capabilities are implemented in the system for the purpose
of identification of opportunities and its utilization [29].
8 1 Introduction
1.3 Opportunistic Spectrum Usage
The allocation of spectrum in Singapore is shown in Fig. 1.2, and it might be noted
that almost all frequency bands up to 30 GHz have been allocated. Even though
all frequency bands have been allocated, they are under-utilized over a given space
and time. A recent study conducted by researchers in Singapore has shown that the
average utilization of spectrum in 80 MHz–5.85 GHz range is only 4.54 % [30]. The
average spectrum occupancy in Singapore is shown in Fig. 1.3.
Spectrum occupancy for different locations all over world have been collected and
studied by various researchers. Shared Spectrum Company is one that is involved
in thorough and wide range of spectrum occupancy studies. The firm has generated
many spectrum occupancy reports in the range of 30 MHz–3 GHz that are publicly
available [31–39]. A collective study of these reports show that on average at any
given place or time, the average occupancy of a particular band of spectrum varies
from 0 to 30 %. Various other researchers have shown similar results in the context
of average spectrum occupancy across different parts of the world [40–47].
These studies show that there exists a paradox between the scarcity of spectrum
and its under-utilization. Under utilization of frequency spectrum implies that white
spaces exist in the spectrum; which can be opportunistically utilized by an intelligent
opportunistic device [48]. The design of such opportunistic spectrum access follows
the principle of overlay spectrum access. Therefore such opportunistic users exist
5000-5850 MHz
4200-5000 MHz
3400-4200 MHz
2700-3400 MHz
2400-2700 MHz
1880-2400 MHz
1710-1880 MHz
1525-1710 MHz
1429-1525 MHz
960-1429 MHz
890-960 MHz
824-890 MHz
790-824 MHz
614-790 MHz
490-614 MHz
406-490 MHz
230-406 MHz
174-230 MHz
80-174 MHz
80-5850 MHz
0 10 20 30 40 50 60 70 80 90 100
Average Occupancy Percentage
Fig. 1.3 Average spectrum occupancy in Singapore

1.3 Opportunistic Spectrum Usage 9
with licensed users in a hierarchical way, where incumbent users (licensed users)
are designated as primary users (PU) while opportunistic users are designated as
secondary users [3]. A secondary user (SU) periodically senses a channel for the
presence of an incumbent. It performs its own communication only when the channel
is free from any incumbent and relinquishes it immediately whenever an incumbent
signal is detected on it. Therefore the interference from secondary users are kept
minimal such that primary users are oblivious of their presence. The design of such
secondary network allows effective utilization of spectrum, but requires cognitive
capability to automatically detect spectral holes in the spectrum and utilize them on
transparent basis [29]. This requirement is the key concept of Cognitive Radio.
A recent trend in wireless communication is towards realization of fully pro-
grammable wireless communication system through computer software. This allows
cellular telephony network operators to cope with the ever changing standards of cel-
lular technology. Various professional and research organization have shown interest
in Software Defined Radio (SDR) that not only solves the problem of coexistence
of heterogeneous networks, but also alleviates the problem of inefficient spectrum
utilization. The ongoing projects Wireless World Initiative New Radio (WINNER)
[49, 50] by European Union and NeXt Generation Communication Program (XG)
[51, 52] by Defense Advanced Research Project Agency (DARPA) of United States
are focussing on the development of technologies and frameworks for flexible radio-
interface and network architecture so as to improve the spectrum utilization. The
committee Dynamic Spectrum Access Network (DySPAN) [53] is one of the stan-
dards committee sponsored by IEEE, aimed at developing standards for radio and
spectrum management. Several working groups of DySPAN have been formulated
to tackle multiple objectives of efficient spectrum utilization.
1.4 Software Defined Radio and Cognitive Radio
Conventional radio communication systems have components implemented in hard-

ware, which restricts the flexibility of the system while reducing the cost of deploy-
ment. A recent change is seen towards the realization of wireless communication
system completely in software. This allows cellular operators to deploy new com-
munication technologies via a software update rather than changing the whole/part-of
radio platform. A Software Defined Radio (SDR) is realization of all transceiver capa-
bility as a software program on an appropriate computing platform (processor) [54].
As an evolutionary step, a SDR may implement band filters on hardware. Most of
the work done on SDR is from the perspective of development of Physical layer of
the Open Systems Interconnection (OSI) protocol stack. But with the requirement
of cross-layer optimization for the purpose of effective communication system, SDR
has evolved with the development of complete protocol stack.
An opportunistic utilization of the spectrum for the purpose of dynamic spec-
trum access will require reconfigurability in waveform and protocol. Therefore an
SDR along with appropriate cognition capabilities can be recognized as a potential
10 1 Introduction
candidate for secondary network operation. The cognitive concept for dynamic spec-
trum access (DSA) was first introduced by Joseph Mitola during his work on SDR
[27, 28, 48]. By definition, a “Cognitive Radio” is a radio that can change its trans-
mitter parameters based on interaction with the environment on which it operates
[3, 5]. A cognitive radio has the capability to develop waveform and generate proto-
cols in real-time using various cognitive capabilities. Basically, it is an enhancement
to SDR, which builds up the reconfigurable radio communication system. FCC has
identified certain features that can be implemented via cognitive radio to effectively
utilize the spectrum [3]:
1. Frequency Agile: The radio utilized by cognitive radio is able to change its
operating frequency so as to effectively utilize the spectrum while mitigating
potential interference to the licensed users.
2. Reconfigurable Modulation: The transmission parameters and generated wave-
form can be reconfigured to adapt to a certain operating environment. This is
required for selecting a robust modulation scheme for the designated environ-
ment which also conforms to the standards defined by regulatory authorities for
shared spectrum usage.
3. Transmission Power Control: The radio can dynamically adjust its transmission
power from maximum allowed EIRP to low power transmission. The operation
at maximum allowed EIRP allows effective and robust communication, while
operating at low transmission power allows improved sharing of the spectrum.
4. Geolocation Awareness: The radio is enabled to identify its location along with
the knowledge of geographical distribution of spectral allocation. This assists in
spectrum sensing to identify spectral opportunities in an effective way. Coop-
erative geolocation awareness of other cognitive radio users allow a radio to
effectively manage the shared spectrum.
5. Mutual Coordination: A coordination based channel sharing can be achieved
between a licensed incumbent and a secondary user. This coordination can be
pre-defined or it can be spawned on a real-time basis.
Since a cognition enabled software defined radio follows open spectrum sharing
policy, it can share spectrum with licensed spectrum or unlicensed spectrum users.
The former type of sharing is called as vertical sharing while latter as horizontal shar-
ing [55]. Therefore vertical sharing refers to the overlay spectrum sharing of licensed
spectrum with primary incumbents. Similarly, horizontal sharing refers to the over-
lay spectrum sharing of unlicensed channels with Short Range Devices (SRDs). A
horizontal sharing may also refer to the sharing of unlicensed spectrum by two or
more heterogeneous cognitive radio user.
1.4.1 IEEE Groups Working on Spectrum Sharing
Over the years, several IEEE working groups have been formulated for the purpose
of effective spectrum sharing. These groups are designated to increase the utilization
of spectrum for the mentioned communication technologies.
1.4 Software Defined Radio and Cognitive Radio 11
1. 802.11a: This group is responsible for designing Dynamic Frequency Selection

(DFS) scheme so as to allow unlicensed users (IEEE 802.11a) to share spectrum on
an overlay basis with primary incumbents such as radar stations [56]. According to
the standards defined, an 802.11a device while operating on a U-NII band, should
perform periodic scan of channels to mitigate interference with radar systems.
2. 802.11k: This group is intended to provide standards for measuring spectral
opportunity while providing means for measurement, reporting, estimation and
characterization of spectrum usage [57]. The measurement values can be utilized
by an opportunistic user.
3. 802.16.2: This group is responsible for mitigating interference and providing
coexistence capabilities to Wireless Metropolitan Area Network (WMAN) sys-
tems based on IEEE 802.16 standards which operate on licensed bands [58].
4. 802.16h: Similar to IEEE 802.16.2, this group is responsible for mitigating inter-
ference and providing coexistence capabilities to license-exempt systems based
on IEEE 802.16 standards [58]. The standard follows the approach of hierarchical
overlay sharing of spectrum by a secondary user.
5. 802.19: This group is known as Coexistence Technical Advisory Group, which
is responsible for designing policies for coexistence between different standards
of communication [58]. Therefore to evaluate the coexistence capability of any
new upcoming radio communication standard, the group can provide analysis to
check its conformance to the policies defined in IEEE 802.19 standards.
6. 802.22: This group is responsible for standardization of wireless broadband inter-
net services via vertical spectrum sharing with TV bands in the range of 54 MHz–
862 MHz [59]. This service is given the name of Wireless Regional Area Net-
work (WRAN), and it utilizes overlay spectrum sharing with white spaces in TV
bands. Since operation of cable and fiber services for broadband access is not
economically viable in rural/sparsely-populated areas, WRAN is developed as an
alternative.
1.4.2 Cognition Cycle
Since a cognitive radio is based on its capability to sense the external environment
and act accordingly, a form of cognition is required for its realization. The cog-
nition cycle comprises of various states through which a cognitive radio cycles to
continuously observe the environment and decide on the best course of action [5].
The observation of environment indicates detection of primary incumbents on differ-
ent spectrum, while actions indicates making a decision on selection of a particular
spectrum band for transmission [28]. This decision also include selecting modulation
scheme, transmission frequency, power, rate etc. An analysis is done before making
a spectrum decision, which comprises of different environment and network parame-
ters, including channel noise, transmission power, number of primary and secondary
users, spectral bandwidth, bandwidth required, SNR required, etc. A cognition cycle
is shown in Fig. 1.4.
12 1 Introduction
Fig. 1.4 Cognition cycle for

cognitive radio
idle
Adapt
Decision Radio sense

Environment
Analyze
The cognitive radio starts from the idle state and goes into environmental sensing
state. During this state, it records all the environment parameters it senses. It analyzes
these environment parameters in the next phase to get more logical and physical fig-
ures. Based on the obtained radio environment parameter and the required operating
conditions (required SNR, bandwidth etc.), the cognitive radio makes a decision in
the next state. The decision is to either adapt the cognitive radio system to the desired
spectrum band with operating behavior and characteristics (center frequency, band-
width, tx power, rx gain etc.) or to go into idle state. The adaptation state is applicable
for both, cognitive radio operating in transmitting or receiving mode. If no suitable
spectrum is found, then the cognitive radio goes into idle mode (sleep mode) and
waits for next sensing cycle.
1.4.3 Cognitive Engine and Framework
A cognitive engine (CE) is formulated and implemented on various SDR boards in

[29] while utilizing following components:
1. Perception: Sensing done by sensors collect the internal and external data. Inter-
nal data consists of resources available like waveform capabilities, computational
power, battery level etc. External data consists of channel availability, radio reg-
ulations and user demands.
2. Comprehension: Part of the system that comprehends and combines the knowl-
edge along with the sensed data so as to aid adaptation mechanism.
3. Execution: Optimization and adaptation mechanism to control the radio behavior.
Fig. 1.5 A typical cognitive

engine User Domain
Cognitive Engine
Networking and
Radio
Communication
System
Policy Engine
Environment and RF
Policy Domain
channel
A cognitive engine, as shown in Fig. 1.5, is a separate entity that works alongside
the normal communication path. It only tells how to best control the communication
path. User domain conveys the information about the performance requirements of
services and applications (QoS) running on top of the system to the CE. External envi-
ronment and RF channel provide the environmental context to the radio transmission
and reception behavior. For example a multipath environment requires a complex
receiver than a single line-of-sight propagation model. The external environment
information also helps in providing optimization boundaries on the decision-making
and waveform development. Policy domain restricts the system to work within the
working boundaries set by the regulatory authorities like FCC. These policies might
restrict the usage of particular spectrum, the maximum amount of power of a radio etc.
A typical architecture of cognitive engine developed by Virginia Tech’s Centre for
Wireless Telecommunication is shown in Fig. 1.6 [29]. It comprises of a cognitive
controller attached to components like sensors, optimizer, decision maker, policy
engine, radio framework and user interface. The Cognitive controller acts as the
system kernel for the architecture and schedules the input/output and timing of the
other attached components which are discussed below:
14 1 Introduction
User Interface
Sensors
Optimization
Cognitive Controller
Decision Maker
Policy
Verification
Knowledge Base
SQL Database
Radio Framework
Fig. 1.6 Cognitive controller and its interaction with other components
Sensors: Sensors are used to collect the data from external and internal environ-
ments. External environment includes channel noise/availability. Internal environ-
ment includes internal system information like noise power, signal power, bit error
rate and battery life. The information from the sensor component can be used by an
optimization process or a decision based process.
Optimizer: The optimization process is used to maximize the performance of the
system by selecting the most appropriate mode of communication. The parameters
required in the optimization process are user QoS requirement, physical environment,
computational resources available and radio regulations. Given a required QoS, the
optimizer selects an appropriate method of communication which satisfies with the
other parameters.
Decision Maker: This component receives the information from the sensors and
makes decision on the basis of the information. It also decides when a reconfiguration
is needed and whether an optimizer is required, and if required what are its goals. It
can also use past information to help the optimization process. In general, case based
decision theory is used for decision making purposes.
Policy Engine: The waveform selected/created by the optimization process is
checked to see whether it is legal to be used in that location over the given time.
The policy engine consists of a database of the regulatory policies. Whenever a
waveform is selected/generated, it is checked against the database that it is within
the bound. The policy engine is dynamic and evolves with the time to accommodate
any change done by the regulatory authorities.
Radio Framework: Radio framework acts as a middleware in between cognitive

engine and the radio platform. It translates the generic information into the imple-
mentation specific information. All the information such as frequency, modulation
type, bit rate, packet size, are passed to the radio platform using a simple generic
representation with the help of a radio framework.
User Interface: The user interface acts as a control center for the whole cognitive
engine. From this interface a user can control the engine as well as provide parameters
for the working of the system.
Cognitive Controller: Cognitive controller acts as a scheduler and a system ker-
nel to schedule and handle various components attached to it. It utilizes different
implementation capabilities of the components attached to it.
A work on the development of a framework for cognitive radio from the per-
spective of media access control is reported in [60]. The authors developed a MAC
framework for providing composition of different MAC protocols best suited for
a given application, available radio and given spectrum regulatory policies. MAC
protocols are decomposed into their basic functional components and are wired on
the fly to realize envisioned MAC protocol. MAC functionalities are implemented
as a library so that different types of MAC protocols can be envisioned by con-
necting different modules of the library in an appropriate manner. A wiring engine
is used to bind the blocks together and hence provide control and data information
among blocks. Another work on MAC framework development is done in [61] where
authors proposed an Intermittently Connected Cognitive Radio MAC (ICR-MAC)
framework. The framework provides support for cluster formation of the intermit-
tently connected network and tries to maximize the transmission opportunities. The
ICR-MAC framework supports multiple flow-scheduling, i.e. within a cluster more
than one channel can be utilized for communication purposes.
1.4.4 Cognitive Radio Network
The network formation by cognitive radio system can be classified into two types:
Infrastructure based CR network and Cognitive radio based ad-hoc network
(CRAHN) [5, 59]. An infrastructure based CR network has a central network entity
such as an access-point (AP) or a base station (BS), whereas CRAHN does not require
any such central network entity, and therefore users in CRAHN communicate with
each other on ad-hoc basis.
Users in an infrastructure based CR network follow the rules and etiquette set by
the central network entity which also acts as a controller. The users of such network
observe and analyze the spectrum behaviour and transmit this information to the base-
station, so that it can make an appropriate decision. Based on the decision of BS, users
may have to reconfigure their communication parameters so as to remain effective
in the CR network. Figure 1.7a shows 4 network elements along with an access
point. The access point may also act as a gateway for external world communication
(internet).
16 1 Introduction
(a) (b)
Fig. 1.7 Different types of CR networks: a Infrastructure based CR Network. b Cognitive Radio
based ad-hoc network (CRAHN)
Users in CRAHN follow the standards and etiquettes defined for the formation
of the ad-hoc network. Therefore all users in such a network cooperate with each
other to achieve a common objective. Users in such network show collaborative
behavior and exchange the information about the spectrum behaviour which they
learn through observation. If there is any reconfigurability required, all users in the
network have to agree on it. This can be done either by having a voting system, or
through a predefined etiquette which is followed by all users in the network. Users
in such a network, if within communication range, directly communicate with each
other. If users are not within communication range, then the message passing is done
through multi-hopping. In multi-hopping, the message traverses through the sender to
destination with the help of intermediate users (part of routing scheme). Figure 1.7b
shows a CRAHN operation with 4 users in the network. The dotted line in the figure
indicates the routing required in the absence of a direct communication path or when
a direct communication path is not economically suitable.
1.5 Quality of Service (QoS)
In a traditional wireless network, all devices are given equal access and priority.
If the total demand of bandwidth resource from the traffic generated by all users
in network exceeds the available bandwidth, then the throughput of traffic from all
users is equally impeded [14, 62, 63]. This has varying impact on user experience
depending upon the type of application. For instance, a second delay of page loading
in web browsing application will not affect user experience; while the same delay
experienced by a VoIP user will result in inconvenience. Therefore to avoid incon-
venience to the user, a certain guarantee of resource is provided to the application.
The guarantee can be provided in terms of throughput, latency, packet drop rate, and
service response time. This is defined by Quality of Service (QoS), which defines
the ability of a system to guarantee an assured grade of service. To achieve a defined
QoS in a system with static resource, packets are prioritized in order of their user
experience requirements. For instance, a VoIP packet will have high priority over an
HTTP (web page) packet. A higher priority packet gets higher chance/precedence
of packet transmission over low priority packets. Therefore when the total demand
1.5 Quality of Service (QoS) 17
of bandwidth resource from the traffic generated by the users in network exceeds
the available bandwidth, then a higher priority packet is given the precedence for
transmission, while blocking low priority packets. This results in effective band-
width utilization as a result of low contention and hence low packet collision [13].
Also a high priority packet gets the transmission opportunity thereby maintaining the
required QoS. All of the above is realized with the help of blocking of low priority
packets, which can tolerate certain amount of delay. The blocked packets are released
when the transmission of higher priority packet is completed.
A related term in the context of QoS is Quality of Experience (QoE). QoE is a
subjective term and is used to describe the experience/satisfaction of a user with a
particular service/system [15]. QoE and QoS are mutually dependent on each other
and in order to achieve a particular level of QoE, some bound on QoS must be
guaranteed. A typical subset of QoS in telecommunication is defined as Grade of
Service (GoS). GoS is used to define the blocking/dropping probability of the calls
made in a system. Generally GoS is calculated by taking the ratio of total number of
calls dropped or blocked to the total number of call attempts made.
QoS provisioning is of crucial importance for network supporting real-time traf-
fic. For instance, in a health-care monitoring system, the results from various sensors
are aggregated and transmitted on a real-time basis. This allows servicing any emer-
gency/urgency required by a patient immediately at the time of action. If real-time
traffic is multiplexed with non-real time traffic, packet prioritization or QoS provi-
sioning is required [64]. To implement QoS provisioning following techniques need
to be implemented by a system:
1. Access Category: To provide packet priority, packets need to be classified into

different categories. These categories are termed as Access Category (AC). An
access category refers to the priority category of a packet to access a resource.
The WMM profile in IEEE 802.11e defines 4 types of access category [14].
2. Traffic Classifier: To identify the access category (AC) of a packet, some kind
of identification scheme is required. In general, packet marking is required to
identify a particular type of data packet, which is in general done by application.
But if an application does not support packet marking scheme, then a traffic
classifier is required to classify a packet [65].
3. Packet Admission Control: Once the packets are identified, a high priority packet
is allowed to access the resource while a low priority packet is blocked. This is
achieved via packet admission control technique, which follows certain rule to
guide packet admission, while also mitigating the problem of low priority packet
starvation [66].
4. Starvation Prevention: If low priority packets in a system are always blocked
due to excessive high priority packets in the network, this results in resource
starvation of low priority packets [67]. Therefore the user experience for low
priority packets may degrade, resulting in low QoS of the system. Therefore a
scheme is required to prevent starvation of low priority packet while maintaining
a graceful degradation of QoS.
18 1 Introduction
5. Resource Requirement Identification: A packet in general may require a high

priority in the context of either latency or throughput. This is discussed in Chap. 3,
Sect. 3.1.1, where resource requirement identification of a packet is done on the
basis of its characteristics. If a packet requires high priority in the context of
latency, it implies that it gets the first access to the available resource than its
counterpart packets [68]. On the other hand, if a packet requires high priority in the
context of throughput, it implies that the packet is transmitted over a non-shareable
resource which has enough bandwidth to satisfy its throughput requirement [69].
If the packet is transmitted over a shareable resource, a reservation (guarantee) is
done to make sure the packet receive enough bandwidth to satisfy its throughput
requirement.
As discussed above, resource identification is done to determine the kind of

resource requirement needed by a packet so as to maintain acceptable user expe-
rience. The following sections discuss on how to implement QoS provisioning in the
context of different resources.
1.5.1 QoS Provisioning for Latency Guarantee
Traditional IPv4 network supports a level of prioritization to traffic using Type of

Service (ToS) field in IP header which is replaced by Differentiated Service Code
Point (DSCP) field. This implementation is termed as Differentiated Services (Diff-
Serv) [12, 13]; which provides low-latency services to time-critical (real-time) traffic
while coexisting with best effort traffic. This is implemented at the Layer-3 of the
protocol stack and hence is able to define packet-forwarding (routing) behavior via
various DiffServ classes. DiffServ defines the Assured Behavior with four separate
classes, each having three categories for drop precedence (Low/Med/High) of pack-
ets. The categorization into separate classes and drop precedence allows the DiffServ
to control the routing and retransmission delay of individual packets. Therefore in a
congested network, a router enabled with appropriate DiffServ will drop a low prior-
ity packet as defined in the policies, so as to allow a high priority packet to access the
resource. This mechanism also allows selection of queuing algorithms like: Priority
Queuing, Fair Queuing and Weighted Fair Queuing (WFQ) [70] allowing a network
operator to control the behavior of queuing delay of packets.
In wireless networks, IEEE 802.11e working group is responsible for the provi-
sioning of QoS in IEEE 802.11 networks [71]. It defines enhancement for the prior-
itization of packets that are targeted at MAC layer of wireless network. The basis of
QoS prioritization is the introduction of new coordination function: Hybrid Coordi-
nation Function (HCF). HCF utilizes two method of channel access: HCF Controlled
Channel Access (HCCA) and Enhanced Distributed Channel Access (EDCA) [72].
HCCA defines the central coordination based contention-free access of channels,
while EDCA defines the distributed contention based access of media. Therefore
HCF has the capacity to divide a MAC superframe into two separate durations:
1.5 Quality of Service (QoS) 19
Table 1.1 Contention window size of different access category [72]

AC Type CWmin CWmax
aCWmin +1 aCWmin +1
AC_VO Voice 4 −1 2 −1
aCWmin +1
AC_VI Video 2 −1 aCWmin
AC_BE Best effort aCWmin aCWmax
AC_BK Background aCWmin aCWmax
Table 1.2 Typical values of EDCA parameters for aCWmin = 15 and aCWmax = 1023
AC AIFSN Default CWmin Default CWmax
AC_VO 2 3 7
AC_VI 2 7 15
AC_BE 3 15 1023
AC_BK 7 15 1023
Contention free period to be utilized by HCCA and Contention period to be utilized

by EDCA. IEEE 802.11e defines four access categories for packet prioritization.
The packets are prioritized by defining different arbitration inter-frame spaces and
contention window size. An arbitration inter-frame space (AIFS) refers to the time
duration which a user has to wait before transmitting a packet. A shorter AIFS gets
higher probability of transmission and vice-versa. The value of AIFS is obtained
with the help of Arbitration Inter-frame Space Number (AIFSN) and is calculated
as [72]:
AIFS[AC] = AIFSN[AC] ∗ aSlotTime + SIFS (1.1)
Here aSlotTime is the size of time-slot of the system time, while SIFS is the size of
short inter-frame space defined for highest priority packets like acknowledgement
(ACK). Table 1.1 shows the maximum (CWmax ) and minimum (CWmin ) contention
window size of different access category of packets defined for a typical aCWmax
(maximum) and aCWmin (minimum) values utilized by the system respectively.
For a typical value of aCWmin = 15 and aCWmax = 1023 in 802.11e PHY,
Table 1.2 shows the default contention window parameters of EDCA.
A shorter value of AIFS allows a high priority packet to win the arbitration period
and gain the opportunity for transmission. Apart from contention based data transfer,
IEEE 802.11e also provides contention-free access of channel by utilizing a period
called as Transmission Opportunity (TXOP). A TXOP is a maximum bounded time
interval during which a user may send as many packets as it intend to. The TXOP is
only defined for voice (AC_VO) and video (AC_VI) data packets.
Wi-Fi Multimedia (WMM) [14], is an alliance for certifying interoperability of
devices based on IEEE 802.11e standard. It defines the standard of access categories
for packet prioritization, however it lacks the standard for providing guaranteed
throughput. Therefore throughput guarantee is a separate issue that needs to be taken
care of.
20 1 Introduction
1.5.2 QoS Provisioning for Throughput Guarantee
Throughput guarantee is separately handled by a system providing QoS. Throughput

may indirectly affect the latency of a traffic from a user [73]. For instance, if the
throughput requirement of a traffic exceeds the available bandwidth resource, then
the packets need to queue which increases the latency of the system. In a video-
on-demand system, if the communication media is insufficient to support the data
traffic, then the video playback suffers from frequent buffering. In a packet-switching
network, a guarantee on assured level of throughput can be provided via: Resource
Allocation and Packet Admission Control [66, 74].
1.5.2.1 Resource Allocation
Resource allocation refers to the allocation of wireless resources to all the users in a
system [75]. A resource allocation scheme may provide a guarantee on what kind of
resource is allocated to a particular user. A user with high demand for throughput may
be allocated a high bandwidth resource. For instance, a user who frequently watches
high-definition content on a wireless device may be allocated a high bandwidth
channel to support the traffic throughput. In a multi-channel environment, resource
allocation translates to allocation of channels to users. Therefore a user may be
allocated with one or more than one channel to accommodate his need. There can be
multiple objectives of a resource allocation scheme including maximizing the system
throughput or providing fairness among users in the network.
1.5.2.2 Packet Admission Control
In a resource constrained environment where resources cannot be allocated or dis-

tributed, for instance a network operating on a single channel, the available resource
is reserved by having packet admission control scheme. A packet admission control
scheme allows a higher priority packet to access the resource, while a low priority
packet gets blocked [76]. Therefore this ensures that the resource is entirely utilized
by high priority packets and hence is guaranteed to be available.
1.6 Channel Selection Techniques in Cognitive

Radio Network
The performance of a network depends upon the bandwidth and quality (SINR) of
the wireless spectrum (channel) onto which it operates. For cognitive operation of a
system on open spectrum sharing basis where channel availability sporadically varies,
selection of channels play an important role in determining the quality of service of
1.6 Channel Selection Techniques in Cognitive Radio Network 21
network operation [77]. For instance, selecting a channel whose availability varies
frequently will cause network to switch channels frequently resulting in jitter and
increase in packet loss. Therefore channel selection is an important task that need to
be performed by spectrum manager of the cognitive radio system while considering
various parameters like channel availability, SINR, bandwidth, path loss and primary
user behavior.
Spectrum decision is a three step process: spectrum characterization, spectrum
selection and system reconfiguration [78]. Spectrum characterization involves iden-
tifying different characteristics of spectrum from local and statistical observation
including: path loss, error rate, interference, switching delay (hardware dependent)
and capacity. Based on regulatory policies additional characteristics like maximum
holding time, allowed bandwidth and allowed maximum EIRP need to determined.
Spectrum selection involves selection of appropriate spectrum based on the charac-
teristic obtained and the demand of QoS of secondary users. System reconfiguration
involves adapting transmission parameters of the transceiver so as to operate within
the selected band of spectrum.
There have been several attempts in developing efficient channel selection algo-
rithms which provide a channel or a set of channels satisfying certain constraints
[3, 78]. These constraints are normally applied on the characteristics of the given
spectrum. Spectrum decision can be classified on the basis of network mode of opera-
tion. In infrastructure network mode of operation, channel selection decision is made
by the central network entity: base-station or access point. Based on the network eti-
quettes, single or multiple channels can be selected by the base-station and allocated
to different end-users in the network. In ad-hoc network mode of operation, chan-
nel selection decision is made by individual users on collaborative behavior basis.
Each communicating pair in ad-hoc network selects a channel which is orthogonal
to the channels selected by other communicating pairs. This allows multiple pairs to
simultaneously perform communication without interference.
1.6.1 Channel Selection in CR Based Infrastructure Network
As discussed in Sect. 1.4.1, IEEE 802.22 WRAN standard has been formulated for
the purpose of broadband provisioning in rural area via vertical spectrum sharing on
TV bands with the help of cognitive system. In [8], dynamic channel management
for IEEE 802.22 has been discussed from the perspective of protection of incumbents
from secondary network activity. The work describes the operation of channel sens-
ing, channel selection and channel information maintenance for dynamic channel
management. Since WRAN operates on TV white space which might operate with
a bandwidth of 6 or 8 MHz (depending on region), the system is able to adapt its
bandwidth. A channel width adaptation algorithm is presented in [79], which can
maximize throughput or minimize the power consumption. A detailed analysis is
done to show the impact of channel width on throughput, range and power con-
sumption. An adaptive spectrum assignment algorithm is discussed in [80], which
22 1 Introduction
handles the spectrum variation and fragmentation. The technique is called signal
interpretation before Fourier transform and it reduces the time to detect transmission
in variable spectral width system by analyzing raw signals in time-domain. In [81],
a game theoretic framework has been discussed, which models the channel selection
as a non-cooperative game. Spectrum bandwidth, interference and spectrum oppor-
tunity holding time is utilized for the purpose of defining the cost function. Cost
minimization is done and the spectrum with the minimum cost is selected as the
best spectrum for the secondary user. A pre-defined policy based channel selection
framework is discussed in [9]. The framework utilizes multiple criteria including
minimum SINR, average SINR, coverage area, occupancy history, number of sup-
portable nodes and incumbent appearance probability to select multiple channels.
Three different schemes have been discussed for channel selection depending upon
the weights assigned to the criteria, selection order and a combination of these. In [82],
channel switching delay is taken into consideration for channel selection. Various
techniques including geolocation historical information is utilized to avoid selecting
previously occupied channels. A trade off between higher throughput (bandwidth)
channels and switching overhead is studied in [10]. The results show that utiliz-
ing higher throughput channels (with large bandwidth) with few secondary users
increases the switching overhead due to increase in sensing time.
1.6.2 Channel Selection in CR Based Ad-hoc Network
A centralized and distributed protocol for spectrum allocation is discussed in [83].

The concept of time-spectrum block is introduced to model the spectrum reservation
by a secondary user. Here a portion of spectrum is reserved for a particular time to
a user over a given region, such that the communication does not interfere with the
incumbents. In [64], channel selection based on switching techniques is discussed.
Two channel switching techniques are proposed in case the current channel becomes
unavailable: periodic switching and triggered switching. In periodic switching, a CR
user switches to a new channel at the start of pre-defined interval, while in triggered
switching a CR user immediately switches to a new channel whenever the current
channel becomes unavailable. In [84], the authors have proposed a game theoretic
based channel selection scheme for cognitive radio based ad-hoc networks. Two
different objective functions of game have been proposed to capture the utility of
selfish and cooperative users. Utility function for selfish user accounts for interfer-
ence on a given channel, while utility function for cooperative user accounts for
interference on a given channel and the interference it will cause to other users. A
potential game has been formulated with the help of objective functions and is shown
to converge at Nash equilibrium point. A learning automaton based channel selection
algorithm is proposed in [85]. The algorithm optimally selects a channel which max-
imizes the probability of successful transmission and reduces the number of costly
channel switching. In [86], the authors propose a metric to characterize the channel
quality and use it to select the channel. The Busy Time Ratio (BTR) metric uses
1.6 Channel Selection Techniques in Cognitive Radio Network 23
collaborative sensing and thus reduces the sensing overhead by allowing each CR
user to scan only the channel in which they were already performing some com-
munication. In [87], an automatic channel selection algorithm has been proposed
which co-exists with a Wi-Fi network. The algorithm selects a channel based on
measured utilization of the ISM band while also considering the QoS requirement
of the CR users. The algorithm can also achieve user selectable trade-off between
system throughput and users’ fairness. In [88], the author compares the multi armed
bandit (MAB) problem with that of the channel selection in cognitive radio. The
author reviews and performs comparisons of algorithms for finite time MAB in sta-
tic and dynamic environment. A QoS based framework is discussed in [89] which
utilizes the bandwidth delay product as a measure for selecting the optimal channel
so as to meet the user’s QoS requirement. In [90], the authors discussed a distributed
channel selection protocol for single hop transmission using two transceivers and a
common control channel. The protocol facilitates load balancing to the system and
thus decreases cluttering on the same channel. Graph coloring based resource allo-
cation techniques can be utilized for allocating channels among users of CR based
ad-hoc network. The graph coloring based resource allocation scheme are discussed
from the perspective of self-coexistence in Sect. 1.8.
In case of multiple cognitive radio network operation in a given region, the channel
selection scheme needs to mitigate the channel selection collision. Channel selection
collision implies selection of the same channel by two or more different networks
which may result in access collision. This is discussed in more detail in Sect. 1.8.
1.7 MAC Protocols for Cognitive Radio Networks
After a channel or a set of channels has been selected by a user in cognitive radio
network, users perform communication on the selected channel. If two or more
users in the vicinity utilize the same channel for communication, it results in access
collision as shown in Fig. 1.8 [91]. This can be mitigated with the help of an access
etiquette which needs to be followed by all users in the network so that frequency-
temporal usage among all user pairs are orthogonal to each other in a given region.
The technique is termed as media access control (MAC). This can be achieved
via random access schemes or time-slotted schemes. The MAC scheme utilized by
cognitive radio network differs from traditional wireless network in the sense that it
incorporates two major design factors: quiet-period and multi-channel environment
[11]. Multi-channel environment allows the operation of MAC scheme to switch to
another channel instantly whenever a primary incumbent returns back to the current
channel. Quiet-periods are utilized for the purpose of spectrum sensing and are
discussed in detail in Sect. 3.4.2.
In a cognitive radio environment, users can choose any of the available channels
and hence sender and receiver need to agree upon a common communicating channel
which is available to both. For this purpose, initial handshake signals are required to
negotiate the communication channel. These signals are termed as control signals,
24 1 Introduction
l3 node 2
ne
node 1 an
Ch
Access Collision
node 3
node 4
Channel 3
Fig. 1.8 A network of nodes operating in a given region. Nodes 1 and 4 are trying to communicate
with nodes 2 and 3 on the same channel, which results in access collision
and since control signals are exchanged between sender and receiver, the design
constraint requires them to be on the same signalling channel [92]. This channel
is termed as common control channel, which additionally increases the resource
requirement of the network by having an additional channel for the purpose of con-
trol signalling. This is termed as common control channel problem. Therefore the
resource complexity of a MAC scheme depends upon whether it utilizes a control
channel or not. An additional resource complexity can be defined in terms of hard-
ware requirement of the MAC scheme: i.e. number of transceivers required. Some
MAC schemes utilize single transceiver while others require the usage of multiple
transceiver [11]. The MAC scheme differs for different network mode of operation
and hence infrastructure based CR network and CRAHN will employ different MAC
schemes.
1.7.1 Random Access Based MAC Scheme
Random access schemes do not require time synchronization and are generally
based on the principles of carrier sense multiple access with collision avoidance
(CSMA/CA). In [91], a random access protocol based on CSMA/CA is discussed
for infrastructure based CR networks. The discussed scheme utilizes one single trans-
ceiver and closely integrates with the PHY layer to control the transmission power and
the throughput of the CR network. Therefore the MAC scheme also supports some
form of coexistence with the primary system following classical CSMA scheme. Mul-
tiple random access schemes have been discussed for ad-hoc cognitive networks. In
[93], a distributed MAC scheme has been discussed for users with hardware con-
straints of single transceiver and limited spectrum sensing capability. The protocol
performs limited fine spectrum sensing and makes a decision on the number of chan-
1.7 MAC Protocols for Cognitive Radio Networks 25
nels that need to be scanned. The protocol utilizes a common control channel for the
purpose of signalling by performing contention on it. The winning pair of contention
then performs channel sensing on the given channels. Based on stopping rule for
spectrum sensing, the pair may perform transmission or start sensing a new channel.
A single transceiver based MAC scheme is also proposed in [94] which utilizes a
method of cross-channel communication, in which the sender and receiver utilize
different spectrum for sending and receiving. Therefore large spectrum block can
be allocated for sending purposes while small spectrum block can be allocated for
acknowledgement purposes. An opportunistic spectrum access (OSA-MAC) scheme
for cognitive radio ad-hoc network is discussed in [95]. The scheme utilizes single
transceiver with a common control channel. All nodes in the network switch to con-
trol channel for the purpose of contention and the winning pair switches to the data
channel for the purpose of communication. Before performing communication, the
data channel is sensed for primary incumbent activity by the communicating pair
during quiet-period. If the channel is sensed busy, the communication is dropped for
the current superframe. A superframe structure of OSA-MAC is shown in Fig. 1.9.
The protocol is able to solve the problem of multi-channel hidden terminal through
temporal synchronization during contention phase.
In [96], the concept of busy-tone is introduced, where a small band of spectrum is
mapped to a data channel. This band of spectrum is termed as busy-tone-band, and is
responsible for identifying whether a data channel corresponding to the busy-tone-
band is occupied or not. Whenever a user occupies a data channel it sends a busy tone
to its corresponding busy-tone-band. A common control channel is utilized for control
signals handshake. Three transceivers are utilized, each assigned on data channel,
Superframe
Common control
contention Data transfer period

channel
Data channels
Time
Quiet period
Fig. 1.9 Superframe structure of OSA-MAC. All nodes switch to the common control channel
during the contention period. The winning node pair move to the agreed data channel and perform
sensing of the channel during quiet-period. Based on the sensing result, the node pair either defer
their communication for the current superframe or go ahead with the communication
26 1 Introduction
control channel and busy-tone-band. This protocol also solves the problem of hidden
terminal. A multi-transceiver based MAC is also discussed in [97], which utilizes a
dedicated transceiver for a dedicated common control channel. The scheme follows
distributed channel assignment by following a list of currently used bands and free
bands. The sender and receiver follow data exchange similar to optional RTS-CTS
IEEE 802.11 standard on a common agreed available channel. In [98], a distributed
cognitive radio MAC protocol has been discussed which provides detection and
protection of incumbent systems around the cognitive radio user pair. The literature
also describes a method for sensing information exchange between neighbor nodes.
In [99], an adaptive MAC protocol for throughput enhancement in cognitive radio
has been discussed which changes its transmission mode between frame recovery
mode and dual transmit mode according to the channel status. It utilizes two half-
duplex transceivers to achieve this goal. The CR MAC protocol proposed in [100]
utilizes multiple channel sensor and CR-enabled transceiver for efficient utilization
of the unused spectrum while avoiding collision with PUs. It closely integrates the
cooperative sequential spectrum sensing along with packet scheduling at MAC layer.
A Markov chain model is discussed to study the network in saturated as well as in
non-saturated condition.
1.7.2 Time-Slotted Based MAC Scheme
The utilization of random access technique simplifies the design of access scheme but
decreases the efficiency of overall throughput of the system. Also the requirement
of contention makes the control channel vulnerable to control channel saturation
problem. With the increase in number of users, the amount of contention on control
channel will increase leading to saturation of contention and hence the data transfer.
Such problems can be removed by designing a MAC scheme which does not involve
contention on a common control channel or by removing contention completely.
Time-slotted MAC protocols remove the contention but require network wide tight
synchronization of time.
IEEE WRAN 802.22 was briefly discussed in Sect. 1.4.1, which is the standard
for infrastructure based cognitive operation over TV white space so as to enable
broadband access in rural areas [59]. A base station coordinates all its users which
are termed as consumer premise equipment (CPE). The MAC scheme is time-slotted,
comprising of superframes, which in turn comprises of multiple MAC frames. Super-
frame control header (SCH) precedes all superframe and consists of network informa-
tion which is distributed to CPEs. Each MAC frame is composed of two subframes:
Downstream (DS) and Upstream (US). A DS subframe consists of a single burst
of packet from a given CPE along with the scheduling information of user bursts.
A US subframe consists of multiple packet bursts, each transmitted from different
CPE. Among others, a US subframe is also used for self-coexistence through the help
of coexistence beacon protocol (CBP). CBP is discussed in Sect. 1.8. The standard
supports two type of spectrum sensing: fast and fine, allowing control over latency
1.7 MAC Protocols for Cognitive Radio Networks 27
and quality of spectrum sensing results respectively. In [101], an improvement to

IEEE 802.22 standard has been proposed. The discussed centralized MAC protocol
for cognitive user coexists with primary user by using in-band/out-band signalling
mechanism and by utilizing extra channels to broadcast current channel list.
In [102], a cognitive MAC protocol for multi-channel wireless (C-MAC) is pro-
posed which utilizes distributed quiet period for primary user (PU) signal detection.
In this protocol each channel is logically divided into recurring superframes com-
prising slotted beaconing period in which users exchange information and performs
channel contention. The concept of rendezvous channel (RC) is introduced in which
users synchronize, coordinate, perform reservation of channels. Users join a network
through the help of beacon period of the superframes of a channel. The concept of dis-
tributed beaconing is utilized for solving the hidden terminal problem, by extending
the beacon through re-broadcast. The utilization of multiple transceiver allows the
MAC scheme to be robust against spectrum changes, as out-of-band sensing is per-
formed on distributed quiet periods. The MAC scheme is targeted towards providing
higher aggregate link throughput.
A combination of time-slotted and random access schemes have emerged as a
new design trend for MAC schemes. Some literatures refer to them as hybrid MAC
protocols [11]. In [103], a single transceiver based MAC scheme is proposed which
utilizes a dedicated common control channel for signaling purposes. The protocol
utilizes pre-defined window period for user coordination over the choice of frequency
spectrum. Users form different cluster, with each cluster member having the intention
to communicate within the cluster. The communication is done by utilizing conven-
tional contention based data transfer. In [104], a control channel-less MAC protocol
has been proposed and referred as Synchronized MAC (SYN-MAC) protocol for
multi-hop cognitive radio networks. In this approach all nodes are assumed to be
equipped with two radios. One radio is used for listening (listening radio) control
signals and the other for both receiving and transmitting data (data radio). Here time
is divided into time-slots, where each slot represents a data channel. These time-
slots are used for control signal exchange, while data exchange can be done on any
agreed channel. Users switch to the channels during the start of the time-slot and
if interested perform control signal exchange. An opportunistic MAC scheme with
the requirement of two transceivers and a dedicated common control channel is pro-
posed in [105]. One radio is tuned to CCC while the other is tuned to data channels.
The data channels are time-slotted while CCC is partly slotted with random access
duration. During the partly slotted duration of CCC, users communicate about the
availability of channels through sensing. During the random access period, users
perform negotiation of channel access through contention. In [106], a distributed
sensing mechanism has been proposed utilizing two radios; one radio is used for
communication purpose while the other one is used for dedicated sensing purpose.
Therefore it allows simultaneous communication and sensing. A dedicated CCC is
utilized for signaling purposes. The time is divided into fixed length beacon intervals
with two phases: Announcement Traffic Indication Message (ATIM) window and
DATA window. Minislot protocol is utilized during ATIM window to learn and dis-
sipate network-wide spectral opportunities. The actual communication is performed
28 1 Introduction
during DATA window. [107, 108] apply the multi-channel and Time Division Mul-
tiple Access (TDMA) technique to increase the throughput of the cognitive radio
network. The system is divided into fixed length time frames comprising of ATIM
window and communication window similar to that discussed in [56]. The commu-
nication window is further divided into fixed size time-slots for the provision of Time
Division Multiple Access. The contention is done during the ATIM window and it
reserves the channel-time slots for actual transmission during the communication
window.
1.8 Self-coexistence in Cognitive Radio Networks
With multiple cognitive radio networks operating in a given region, a node in a

network has to mitigate the harmful effects of interference from licensed users and
nodes of other cognitive radio networks. Interference from other cognitive radio
network involves: different cognitive radio network operating with different set of
protocols (heterogeneous) or different cognitive radio networks operating under the
same set of protocol (homogeneous). Mitigating interference from cognitive radio
network operating under same set of protocol requires self-coexistence techniques.
Self-coexistence techniques are designed so as two or more homogeneous cognitive
radio network can cooperate and coexist with minimal interference.
A three step self-coexistence technique is proposed in [109]. Each of these steps
provide self-coexistence from different perspective. A sequential flow is presented
for making a decision on the selection of technique, which includes parameters like
coexistence complexity, reconfigurability and performance of the selected scheme.
The techniques in the order of selection process are: resource relocation, transmission
power control, and resource sharing.
1. Resource Relocation: Resource in the context of cognitive radio network refers

to a channel or a combination of time-duration with a channel [109]. The first
step in resource relocation is to identify the resources over which the users are
interfering. Once these resources are identified, they are redistributed among the
users such that they do not interfere while also maintaining a certain level of QoS
for their respective network. Resource relocation achieves high performance gain
in comparison to other methods of self-coexistence. Also the complexity of this
method is lower than other methods of self-coexistence.
2. Transmission Power Control (TPC): This technique is performed when resource
relocation based self-coexistence is not feasible. This technique requires adapting
the transmission power of interfered users so as to separate the interference range
of users. TPC technique allows self-coexistence but at the cost of decrease in
communication range and throughput. Therefore an optimization is required to
negotiate between the level of interference and QoS requirement.
1.8 Self-coexistence in Cognitive Radio Networks 29
3. Resource Sharing: When resource relocation and TPC based self-coexistence are
not successful, then the resource is shared in frequency or time domain between
the interfered users. Sharing a spectrum in frequency domain implies dividing a
large block of spectrum into small chunks (fragmentation) and allocating them to
the interfered users [110]. Similarly sharing a spectrum in time-domain implies
sharing it on time division multiplex basis, where a particular time duration is
allocated to a given interfered user [111]. This technique is more complex and
less efficient compared to other techniques of self-coexistence.
1.8.1 Resource Relocation Based Self-coexistence
Various works on resource relocation have been done in wireless networks from the
perspective of self-coexistence [112]. These methods can be applied in cognitive
radio networks with the additional design consideration of sporadically available
channels. In IEEE 802.22 standard, coexistence beacon protocol is utilized for sup-
porting intra-network coexistence [113]. Intra-network communication is done with
the help of CBP [114] where intra-network packets are transmitted during a specifi-
cally defined duration: coexistence window. During this duration, contention based
access is done to access the spectrum with higher priority given to the inter-base
station messages.
Since resource relocation implies resource allocation, graph coloring based chan-
nel allocation schemes can be utilized for the same [115]. The users in the network
are mapped as vertices while interference between them is mapped as edges between
vertices. Each user has a list of available colors (channels). The aim of a graph
coloring scheme is to allocate colors to vertices such that no two neighbor vertices
are allocated with the same color. In infrastructure based network, base stations are
mapped as vertices. In [116], a graph coloring based channel allocation scheme is
proposed which allocates channel to different users in a cognitive radio network
with proportional fairness. A collaborative labeling rule is proposed which guaran-
tees users’ fairness of service capacity. In [117], a utility graph coloring scheme is
proposed which consider allocation of colors based on the weights assigned to them.
The overall aim of the technique is to maximize the overall spectrum utilization.
An enhanced MAC scheme for IEEE 802.22 standard is also proposed to maximize
the spectral efficiency with the help of channel aggregation and fragmentation. In
[118], the channel allocation problem is converted into multi-stage modified minor-
ity game, where the aim is to select a channel free from interference. The number of
competitors is known with the help of foreign beacon period (broadcast beacons of
foreign 802.22 networks) of IEEE 802.22 network. The player in the game makes a
decision whether to stay on the current channel or to move to other channel with low
interference. A channel switching cost function is defined for the same. A dynamic
hopping based self-coexistence technique for IEEE 802.22 networks is proposed in
[114]. A group of IEEE 802.22 networks form a community and hop periodically
in a defined manner from one channel to another. The successive access of channel
30 1 Introduction
by different networks is separated by quiet-period, during which other networks can

perform out-of-band sensing. This increases the performance and reliability of the
system while maintaining coexistence among all IEEE 802.22 networks.
1.8.2 Resource Sharing Based Self-coexistence
Frequency division multiplexing (FDM) based coexistence has been widely studied
over the years for allowing systems to coexist in orthogonal frequency domain [110].
It allows a wide band of spectrum to be sliced into small fragments of spectrum bands,
which can be allocated to individual users. The division of wide frequency band into
small bands results in fragmentation, which has been widely studied for dynamic
spectrum access. In [119, 120], performance analysis of spectrum fragmentation
is done. It is shown that excessive fragmentation increases the spectrum utilization
but decreases the performance of the system. Experimental and analytical model
have been proposed to validate the system. Since frequency division multiplexing
is done at the Layer 1 (PHY) of the protocol stack, it is beyond the scope of this
work. Readers are suggested to go through the literature work done in [121], which
introduces time-frequency-code slicing technique for multi-rate support in networks.
Time division based multiplexing (TDM) has been widely used for resource dis-
tribution [111]. In TDM, the resource access is divided into small time durations,
where all users are allocated with disjoint time-durations to access the given resource.
The computing power distribution among multiple process in a computer system is
an example of TDM system. Conventional Global System for Mobile Communica-
tion (GSM), is based on TDM, where multiple users are assigned the same band but
are separated on the basis of TDM. GSM system utilizes FDM along with TDM to
increase the number of users on the network. For self-coexistence purposes, [109]
utilizes a TDM access (TDMA) based frame sharing approach. Here the superframe
of each network for the same channel is divided in time-domain, such that each
network has access to an exclusive non-overlapping superframe.
1.9 Discussion
The overcrowding of unlicensed bands has resulted in “tragedy of commons,” that

needs to be mitigated by allowing more unlicensed bands for unlicensed users. A
paradox has been shown between the spectrum scarcity and its under-utilization, and
therefore open spectrum access techniques have been envisioned by researchers to
effectively utilize frequency spectrum. To realize the open spectrum sharing tech-
nique, underlay and overlay sharing techniques have been proposed. Underlay com-
munication is useful for wide-band short range devices while overlay sharing is useful
for opportunistic utilization of spectrum with region level transmission range. To real-
ize the implementation of opportunistic utilization of spectrum, cognitive concept
1.9 Discussion 31
of software defined radio is introduced. A cognitive radio is a reconfigurable radio

that can adapt its transmission characteristic according to the environment in which
it operates. Several standards are being developed by various researchers across the
world for the purpose of realization of open spectrum system.
The concept of quality of service from the perspective of throughput and latency
is introduced. Latency guarantee in a network requires prioritization of a packet
from the perspective of access time or queuing delay. While throughput guarantee
in a network requires reservation of a particular resource (bandwidth), so that the
system operates with acceptable grade of service. The performance of a network
depends upon the channel on which it operates. Several research work on single-
channel or multi-channel selection techniques have been introduced while making a
spectrum decision on the basis of various spectrum characteristics. Channel selection
in infrastructure and ad-hoc networks are also discussed, which differ in terms of
distributed and collaborative behavior. To mitigate the access collision on a channel,
MAC schemes have been introduced. These MAC schemes are classified into two
types: random access based or time-slotted based. Random access schemes are gen-
erally utilized in ad-hoc network while time-slotted access schemes are generally uti-
lized in infrastructure based network due to the requirement of tight-synchronization.
To get the best of both classes of access protocols, a new class of MAC scheme has
been introduced which is called as hybrid scheme. To avoid the mutual interference
from other homogeneous CR networks in the vicinity, self-coexistence techniques
have been discussed. Self-coexistence can be done from three perspective: resource
relocation, transmission power control and resource sharing, in the order of increas-
ing complexity and decreasing performance. The different technologies responsible
for individual attribute of QoS can be integrated together to form a framework which
allows QoS provisioning to the whole system.
References
1. Tragos, E., Zeadally, S., Fragkiadakis, A., & Siris, V. (2013). Pectrum assignment in cognitive
radio networks: A comprehensive survey. IEEE Communications Surveys Tutorials, 15, 1108–
1135.
2. Force, S. P. T. (2002). Report of the spectrum rights and responsibilities working group.
Technical Report: Federal Communications Commission.
3. Akyildiz, I. F., Lee, W.-Y., Vuran, M. C., & Mohanty, S. (2006). Next generation/dynamic
spectrum access/cognitive radio wireless networks: A survey. Computer Networks, 50, 2127–
2159.
4. Lee, W.-Y., & Akyldiz, I. (2011). A spectrum decision framework for cognitive radio networks.
IEEE Transactions on Mobile Computing, 10, 161–174.
5. Akyildiz, I. F., Lee, W.-Y., & Chowdhury, K. R. (2009). Crahns: Cognitive radio ad hoc
networks. ad hoc Networks, 7, 810–836.
6. Ghasemi, A., & Sousa, E. (2005). Collaborative spectrum sensing for opportunistic access
in fading environments. In 2005 First IEEE International Symposium on New Frontiers in
Dynamic Spectrum Access Networks, 2005. DySPAN 2005 (pp. 131–136).
7. Hillenbrand, J., Weiss, T., & Jondral, F. (2005). Calculation of detection and false alarm
probabilities in spectrum pooling systems. IEEE Communications Letters, 9, 349–351.
32 1 Introduction
8. Ko, G., Franklin, A., You, S.-J., Pak, J.-S., Song, M.-S., & Kim, C.-J. (2010). Channel man-
agement in IEEE 802.22 wran systems. IEEE Communications Magazine, 48, 88–94.
9. Na, D.-H., Nan, H., & Yoo, S.-J. (2007). Policy-based dynamic channel selection architecture
for cognitive radio networks. In Second International Conference on Communications and
Networking in China, 2007. CHINACOM ’07 (pp. 1190–1194).
10. Xu, D., Jung, E., & Liu, X. (2008). Optimal bandwidth selection in multi-channel cognitive
radio networks: How much is too much?. In 3rd IEEE Symposium on New Frontiers in
11. Cormio, C., & Chowdhury, K. R. (2009). A survey on MAC protocols for cognitive radio
networks. ad hoc Networks, 7(7), 1315–1329.
12. Park, S. (2006). Diffserv quality of service support for multimedia applications in broadband
access networks. In International Conference on Hybrid Information Technology, 2006. ICHIT
’06 (vol. 2, pp. 513–518).
13. Borosa, T., Marsic, B., & Pocuca, S. (2003). QoS support in ip multimedia subsystem using
diffserv. In Proceedings of the 7th International Conference on Telecommunications, 2003.
ConTEL 2003 (vol. 2, pp. 669–672).
14. Wi-fi certified for wmm- support for multimedia applications with quality of service in wi-fi
networks, (2004). Technical Report, Wi-Fi Alliance.
15. Dai, Q. (2011). A survey of quality of experience. Proceedings of the 17th International Con-
ference on Energy-aware Communications, EUNICE’11 (pp. 146–156). Berlin, Heidelberg:
Springer.
16. Fall, K. R., & Stevens, W. R. (2011). TCP/IP illustrated: The protocols (Vol. 1). Boston:
Addison-Wesley.
17. Telecom, I. (1996) Standardization sector of ITU. In Video Coding for Low Bitrate Commu-
nication, Draft ITU-T Recommendation H (vol. 263).
18. Commission, F. C., et al. (2003). Revision of parts 2 and 15 of the commission rules to
permit unlicensed national information infrastructure (U-NII) devices in the 5 GHz band.
Memorandum Opinion and Order, 21.
19. Berger, R. J. (2003). Open spectrum: A path to ubiquitous connectivity. ACM Queue, 1(3),
60–68.
20. Werbach, K. (2002) Open spectrum: The new wireless paradigm. New America Foundation
Spectrum Policy Program.
21. Valletti, T. M. (2001). Spectrum trading. Telecommunications Policy, 25(10), 655–670.
22. Li, H., Shu, T., He, F., & Song, J. B. (2013). Futures market for spectrum trade in wireless
communications: Modeling, pricing and hedging. In 2013 IEEE Global Communications
Conference (GLOBECOM) (pp. 902–907).
23. Spectrum management handbook, (2014). Infocomm Development Authority of Singapore,
Issue 1 Rev 2.4.
24. Hazlett, T. W. (2005). Spectrum tragedies. Yale Journal on Regulation, 22, 242.
25. Kolodzy, P. J. (2006). Interference temperature: a metric for dynamic spectrum utilization.
International Journal of Network Management, 16(2), 103–113.
26. Aiello, G. R., & Rogerson, G. D. (2003). Ultra-wideband wireless systems. IEEE Microwave
Magazine, 4(2), 36–47.
27. Mitola, J., & Maguire, G. Q, Jr. (1999). Cognitive radio: Making software radios more per-
sonal. IEEE Personal Communications, 6, 13–18.
28. Mitola, J. (1999) Cognitive radio for flexible mobile multimedia communications. In 1999
IEEE International Workshop on Mobile Multimedia Communications, 1999. (MoMuC ’99)
(pp. 3–10).
29. Rondeau, T., & Bostian, C. (2009). Artificial Intelligence in Wireless Communications. Artech
House: Artech House mobile communications series.
30. Islam, M., Koh, C., Oh, S. W., Qing, X., Lai, Y., Wang, C. et. al., (2008). Spectrum survey
in singapore: Occupancy measurements and analyses. In 3rd International Conference on
Cognitive Radio Oriented Wireless Networks and Communications, 2008. CrownCom 2008
(pp. 1–7).
References 33
31. Erpek, T., Lofquist, M., & Patton, K. (2007). Spectrum occupancy measurements loring
commerce centre limestone, maine. Shared Spectrum Company Report. Retrieved September
18–20, 2007.
32. Erpek, T., Steadman, K., & Jones, D. (2007). Dublin ireland spectrum occupancy measure-
ments collected. Shared Spectrum Company Report. Retrieved April 16–18, 2007.
33. McHenry, M. A., McCloskey, D., Roberson, D., & MacDonald, J. T. (2005, December).
Spectrum occupancy measurements chicago, illinois. Shared Spectrum Company Report.
Retrieved November 16–18, 2005.
34. McHenry, M. A., & Steadman, K. (2005, August). Spectrum occupancy measurements loca-
tion 1 of 6: Riverbend park, great falls, virginia. Shared Spectrum Company Report. Retrieved
April 7, 2004.
tion 2 of 6: Tyson’s square center, vienna, virginia. Shared Spectrum Company Report.
Retrieved April 9, 2004.
36. McHenry, M. A., & Chunduri, S. (2005, August). Spectrum occupancy measurements location
3 of 6: National science foundation building roof. Shared Spectrum Company Report, Revision,
2. Retrieved April 16, 2004.
37. McHenry, M. A., & McCloskey, D. (2005, August). Spectrum occupancy measurements
location 4 of 6: Republican national convention, New York city. Shared Spectrum Company
Report, Revision, 2. Retrieved New york August 30, 2004–September 3, 2004.
tion 5 of 6: National radio astronomy observatory (NRAO) green bank, west virginia. Shared
Spectrum Company Report, Revision, 3. Retrieved October 10–11, 2004.
39. McHenry, M. A., & Steadman, K. (2005, August). Spectrum occupancy measurements:
Shared spectrum building roof, vienna, virginia. Shared Spectrum Company Report. Retrieved
December 15–16, 2004.
40. Valenta, V., Marslek, R., Baudoin, G., Villegas, M., Suarez, M., & Robert, F. (2010). Survey on
spectrum utilization in Europe: Measurements, analyses and observations. In 2010 Proceed-
ings of the Fifth International Conference on Cognitive Radio Oriented Wireless Networks
Communications (CROWNCOM) (pp. 1–5).
41. Patil, K., Skouby, K., Chandra, A., & Prasad, R. (2011). Spectrum occupancy statistics in
the context of cognitive radio. In 2011 14th International Symposium on Wireless Personal
Multimedia Communications (WPMC) (pp. 1–5).
42. Sanders, F. (1998). Broadband spectrum surveys in denver, co, san diego, ca, and los angeles,
ca: Methodology, analysis, and comparative results. In 1998 IEEE International Symposium
on Electromagnetic Compatibility (vol. 2, pp. 988–993).
43. Chiang, R., Rowe, G., & Sowerby, K. (2007). A quantitative analysis of spectral occupancy
measurements for cognitive radio. In IEEE 65th Vehicular Technology Conference, 2007.
VTC2007-Spring (pp. 3016–3020).
44. Wellens, M., Wu, J., & Mahonen, P. (2007). Evaluation of spectrum occupancy in indoor
and outdoor scenario in the context of cognitive radio. In 2nd International Conference on
Cognitive Radio Oriented Wireless Networks and Communications, 2007. CrownCom 2007
(pp. 420–427).
45. Yin, S., Chen, D., Zhang, Q., Liu, M., & Li, S. (2012). Mining spectrum usage data: A large-
scale spectrum measurement study. IEEE Transactions on Mobile Computing, 11, 1033–1046.
46. Lopez-Benitez, M., Umbert, A., & Casadevall, F. (2009). Evaluation of spectrum occupancy
in spain for cognitive radio applications. In IEEE 69th Vehicular Technology Conference,
2009. VTC Spring 2009 (pp. 1–5).
47. Bao, V. N. Q., Cuong, L. Q., Phu, L. Q., Thuan, T. D., Quy, N. T., & Trung, L. M. (2011).
Vietnam spectrum occupancy measurements and analysis for cognitive radio applications.
In 2011 International Conference on Advanced Technologies for Communications (ATC)
(pp. 135–143).
48. Mitola, J. (2000). Cognitive Radio—An Integrated Agent Architecture for Software Defined
Radio. DTech thesis, Royal Institute of Technology (KTH).
34 1 Introduction
49. Mino Diaz, E., Gelpi, P., von Hafen, J., Hiebinger, L., Jamsa, T., Malmgren, G., et al., (2004).
The winner project: Research for new radio interfaces for better mobile services (Special
section wide band systems). In IEICE Transactions on Fundamentals of Electronics, Com-
munications and Computer Sciences (vol. 87, pp. 2592–2598).
50. Mohr, W. (2005). The winner (wireless world initiative new radio) project-development of a
radio interface for systems beyond 3G. In IEEE 16th International Symposium on Personal,
Indoor and Mobile Radio Communications, 2005. PIMRC 2005 (vol. 2, pp. 857–862).
51. Group, D. X. W. et al., (2003). The XG Vision, Request for Comments, Version 2.0. Prepared
by BBN Technologies Cambridge, Massachusetts.
52. Group, D. X. W. et al., The XG architectural framework. request for comments. Prepared by
BBN Technologies Cambridge, Massachusetts.
53. Harada, H., Alemseged, Y. D., & Holland, O. (2011). IEEE dynamic spectrum access networks
(DYSPAN) standards committee. 2011 2nd International Conference on Wireless Commu-
nication, Vehicular Technology, Information Theory and Aerospace and Electronic Systems
Technology (Wireless VITAE) (pp. 1–5). New York: IEEE.
54. Tuttlebee, W. H. (2003). Software Defined Radio: Enabling Technologies. New Jersey: Wiley.
55. Pollin, S., Ergen, M., Timmers, M., Dejonghe, A., Van der Perre, L., Catthoor, F. et. al., (2006)
Distributed cognitive coexistence of 802.15. 4 with 802.11. In 1st International Conference
on Cognitive Radio Oriented Wireless Networks and Communications, 2006 (pp. 1–5). New
York: IEEE.
56. IEEE 802.11 Working Group, (2012). Wireless Lan Medium Access control (MAC) and
physical layer (PHY) Specifications. IEEE 802.11 Standard.
57. Mangold, S., Zhong, Z., Hiertz, G. R., & Walke, B. (2004). IEEE 802.11e/802.11k wireless lan:
spectrum awareness for distributed resource sharing. Wireless Communications and Mobile
Computing, 4(8), 881–902.
58. Sherman, M., Mody, A., Martinez, R., Rodriguez, C., & Reddy, R. (2008). IEEE standards
supporting cognitive radio and networks, dynamic spectrum access, and coexistence. IEEE
Communications Magazine, 46, 72–79.
59. Cordeiro, C., Challapali, K., Birru, D., & Sai Shankar, N. (2005). IEEE 802.22: the first
worldwide wireless standard based on cognitive radios. In 2005 First IEEE International
Symposium on New Frontiers in Dynamic Spectrum Access Networks, 2005. DySPAN 2005
(pp. 328–337).
60. Ansari, J., Zhang, X., Achtzehn, A., Petrova, M., & Mahonen, P. (2011). A flexible mac
development framework for cognitive radio systems. In 2011 IEEE Wireless Communications
and Networking Conference (WCNC) (pp. 156–161).
61. Wysocki, T., & Jamalipour, A. (2009). Mac framework for intermittently connected cognitive
radio networks. In IEEE 20th International Symposium on Personal, Indoor and Mobile Radio
Communications, 2009 (pp. 481–485).
62. Akin, S., & Gursoy, M. (2011). Performance analysis of cognitive radio systems under QoS
constraints and channel uncertainty. IEEE Transactions on Wireless Communications, 10,
2883–2895.
63. Ishibashi, B., Bouabdallah, N., & Boutaba, R. (2008). QoS performance analysis of cognitive
radio-based virtual wireless networks. In The 27th Conference on Computer Communications.
INFOCOM 2008. New York: IEEE
64. Liang, Z., Feng, S., Zhao, D., & Shen, X. (2011). Delay performance analysis for supporting
real-time traffic in a cognitive radio sensor network. IEEE Transactions on Wireless Commu-
nications, 10, 325–335.
65. Zhang, J., Chen, C., Xiang, Y., Zhou, W., & Vasilakos, A. (2013). An effective network
traffic classification method with unknown flow detection. IEEE Transactions on Network
and Service Management, 10, 133–147.
66. Lee, H.-W., & Chong, S. (2007). Combined packet scheduling and call admission control
with minimum throughput guarantee in wireless networks. IEEE Transactions on Wireless
Communications, 6, 3080–3089.
References 35
67. Pattara-Aukom, W., Banerjee, S., & Krishnamurthy, P. (2002) Starvation prevention and qual-
ity of service in wireless lans. In The 5th International Symposium on Wireless Personal
Multimedia Communications, 2002 (vol. 3, pp. 1078–1082).
68. Sakurai, T., & Vu, H. (2007). Mac access delay of IEEE 802.11 dcf. IEEE Transactions on
Wireless Communications, 6, 1702–1710.
69. Bianchi, G. (1998). Ieee 802.11-saturation throughput analysis. IEEE Communications Let-
ters, 2, 318–320.
70. Ji, L., Arvanitis, T., & Woolley, S. (2003). Fair weighted round robin scheduling scheme for
diffserv networks. Electronics Letters, 39(2), 333–335.
71. Perez-Costa, X., & Camps-Mur, D. (2010). IEEE 802.11e QoS and power saving features
overview and analysis of combined performance (accepted from open call). IEEE Wireless
Communications, 17, 88–96.
72. Choi, S., del Prado, J., Sai Shankar, N., & Mangold, S. (2003). IEEE 802.11 e contention-
based channel access (EDCF) performance evaluation. In IEEE International Conference on
Communications, 2003. ICC ’03 (vol. 2, pp. 1151–1156).
73. Gamal, A. Mammen, J., Prabhakar, B., & Shah, D. (2004). Throughput-delay trade-off in
wireless networks. In Twenty-third Annual Joint Conference of the IEEE Computer and Com-
munications Societies. INFOCOM 2004 (vol. 1, p. 475).
74. Kaneko, M., Popovski, P., & Hayashi, K. (2009). Throughput-guaranteed resource-allocation
algorithms for relay-aided cellular ofdma system. IEEE Transactions on Vehicular Technol-
ogy, 58, 1951–1964.
75. Eryilmaz, A., & Srikant, R. (2007). Fair resource allocation in wireless networks using queue-
length-based scheduling and congestion control. IEEE/ACM Transactions on Networking, 15,
1333–1344.
76. Barry, M., Campbell, M., Veres, A. (2001) Distributed control algorithms for service differen-
tiation in wireless packet networks. In Proceedings of the Twentieth Annual Joint Conference
of the IEEE Computer and Communications Societies. INFOCOM 2001 (vol. 1, pp. 582–590).
New Jersey: IEEE.
77. Joe, I., & Son, S. (2008). Dynamic spectrum allocation mac protocol based on cognitive
radio for QoS support. In Japan-China Joint Workshop on Frontier of Computer Science and
Technology, 2008. FCST ’08 (pp. 24–29).
78. Masonta, M., Mzyece, M., & Ntlatlapa, N. (2013). Spectrum decision in cognitive radio
networks: A survey. IEEE Communications Surveys Tutorials, 15, 1088–1107.
79. Chandra, R., Mahajan, R., Moscibroda, T., Raghavendra, R., & Bahl, P. (2008). A case for
adapting channel width in wireless networks. SIGCOMM Computer Communivation Review,
38, 135–146.
80. Bahl, P., Chandra, R., Moscibroda, T., Murty, R., & Welsh, M. (2009). “White space net-
working with wi-fi like connectivity. Proceedings of the ACM SIGCOMM 2009 Conference
on Data Communication, SIGCOMM ’09 (pp. 27–38). New York: ACM.
81. Malanchini, I., Cesana, M., & Gatti, N. (2009). On spectrum selection games in cognitive
radio networks. In Global Telecommunications Conference, (2009). GLOBECOM 2009 (pp.
1–7). New York: IEEE.
82. Azarfar, A., Frigon, J.-F., & Sanso, B. (2012). Improving the reliability of wireless networks
using cognitive radios. IEEE Communications Surveys Tutorials, 14, 338–354.
83. Yuan, Y., Bahl, P., Chandra, R., Moscibroda, T., & Wu, Y. (2007). Allocating dynamic time-
spectrum blocks in cognitive radio networks. Proceedings of the 8th ACM International Sym-
posium on Mobile ad hoc Networking and Computing, MobiHoc ’07 (pp. 130–139). New
York: ACM.
84. Nie, N., Comaniciu, C. (2005). Adaptive channel allocation spectrum etiquette for cognitive
radio networks. In 2005 First IEEE International Symposium on New Frontiers in Dynamic
Spectrum Access Networks, 2005. DySPAN 2005 (pp. 269–278).
85. Song, Y., Fang, Y., & Zhang, Y. (2007). Stochastic channel selection in cognitive radio net-
works. In Global Telecommunications Conference, 2007. GLOBECOM ’07 (pp. 4878–4882).
New York: IEEE.
36 1 Introduction
86. Yang, G.-H., Zheng, H., Zhao, J., & Li, V.-K. (2006). Adaptive channel selection through
collaborative sensing. In IEEE International Conference on Communications, 2006. ICC ’06
(vol. 8, pp. 3753–3758) .
87. Bennai, M., Sydor, J., & Rahman, M. (2010). Automatic channel selection for cognitive radio
systems. In 2010 IEEE 21st International Symposium on Personal Indoor and Mobile Radio
Communications (PIMRC) (pp. 1831–1835).
88. Kung, L.-H. (2008). Channel selection for cognitive radio terminals, 2008. CS229 Project
Report
89. Talat, S., & Wang, L.-C. (2009). QoS-guaranteed channel selection scheme for cognitive radio
networks with variable channel bandwidths. In International Conference on Communications,
Circuits and Systems, 2009. ICCCAS 2009 (pp. 241–245).
90. Ververidis, C. N., Nasreddine, J., & Mähönen, P. (2009). Distributed channel selection pro-
tocol for single-hop transmissions in cognitive radio networks. Proceedings of the 6th ACM
Symposium on Performance Evaluation of Wireless ad hoc, Sensor, and Ubiquitous Networks,
PE-WASUN ’09 (pp. 155–156). New York: ACM.
91. Lien, S.-Y., Tseng, C.-C., & Chen, K.-C. (2008). Carrier sensing based multiple access pro-
tocols for cognitive radio networks. In IEEE International Conference on Communications,
2008. ICC ’08 (pp. 3208–3214).
92. Lo, B. F. (2011). A survey of common control channel design in cognitive radio networks.
Physical Communication, 4(1), 26–39.
93. Jia, J., Zhang, Q., & Shen, X. (2008). Hc-mac: A hardware-constrained cognitive MAC for
efficient spectrum management. IEEE Journal on Selected Areas in Communications, 26,
106–117.
94. Ma, L., Shen, C.-C., & Ryu, B. (2007). Single-radio adaptive channel algorithm for spectrum
agile wireless ad hoc networks. In 2nd IEEE International Symposium on New Frontiers in
95. Le, L., & Hossain, E., (2008). A mac protocol for opportunistic spectrum access in cognitive
radio networks. In Wireless Communications and Networking Conference, 2008. WCNC 2008
(pp. 1426–1430). New York: IEEE.
96. Ma, L., Han, X., & Shen, C.-C. (2005). Dynamic open spectrum sharing mac protocol for
wireless ad hoc networks. In 2005 First IEEE International Symposium on New Frontiers in
Dynamic Spectrum Access Networks. DySPAN 2005 (pp. 203–213).
97. Pawelczak, P., Venkatesha Prasad, R., Xia, L., Niemegeers, I. G. M. M. (2005). Cognitive radio
emergency networks - requirements and design. In 2005 First IEEE International Symposium
on New Frontiers in Dynamic Spectrum Access Networks. DySPAN 2005 (pp. 601–606).
98. Yoo, S.-J., Nan, H., & Hyon, T.-I. (2009). Dcr-mac: distributed cognitive radio mac protocol
for wireless ad hoc networks. Wireless Communications and Mobile Computing, 9(5), 631–
653.
99. Lee, B., & Rhee, S. H. (2008). Adaptive mac protocol for throughput enhancement in cognitive
radio networks. In International Conference on Information Networking, 2008. ICOIN 2008
(pp. 1–5).
100. Zhang, X., & Su, H. (2011). Cream-mac: Cognitive radio-enabled multi-channel mac pro-
tocol over dynamic spectrum access networks. IEEE Journal of Selected Topics in Signal
Processing, 5, 110–123.
101. Kim, H.-J., Jo, K.-J., Hyon, T.-I., Kim, J.-M., & Yoo, S.-J. (2008). Cognitive Radio Mac
Protocol for Hidden Incumbent System Detection (vol. 5200, pp. 80–89).
102. Cordeiro, C., & Challapali, K. (2007). C-mac: A cognitive mac protocol for multi-channel
wireless networks. In 2nd IEEE International Symposium on New Frontiers in Dynamic
Spectrum Access Networks, 2007. DySPAN 2007 (pp. 147–157).
103. Hamdaoui, B., & Shin, K. (2008). Os-mac: An efficient mac protocol for spectrum-agile
wireless networks. IEEE Transactions on Mobile Computing, 7, 915–930.
104. Kondareddy, Y., & Agrawal, P. (2008). Synchronized mac protocol for multi-hop cogni-
tive radio networks. In IEEE International Conference on Communications, 2008. ICC ’08
(pp. 3198–3202).
References 37
105. Su, H., & Zhang, X. (2007). Opportunistic mac protocols for cognitive radio based wireless
networks. In 41st Annual Conference on Information Sciences and Systems, 2007. CISS ’07
(pp. 363–368).
106. Timmers, M., Pollin, S., Dejonghe, A., Van der Perre, L., & Catthoor, F. (2010). A distributed
multichannel mac protocol for multihop cognitive radio networks. IEEE Transactions on
Vehicular Technology, 59, 446–459.
107. Ke, Y., & Yoo, S.-J. (2009). Mcr-mac: Multi-channel cognitive radio mac protocol for coop-
erative incumbent system protection in wireless ad-hoc network. In First International Con-
ference on Ubiquitous and Future Networks, 2009. ICUFN 2009 (pp. 6–11).
108. Kamruzzaman, S. M., Hamid, M. A., & Abdullah-Al-WADUD, M. (2010). An energy-
efficient mac protocol for QoS provisioning in cognitive radio ad hoc network. Radioengi-
neering, 19, 567–578.
109. Lee, D. H., Jeon, W. S., Jeong, D. G., & Ko, G. (2010). Self-coexistence techniques for
cognitive radio lans/pans. In 2010 IEEE 21st International Symposium on Personal Indoor
and Mobile Radio Communications (PIMRC) (pp. 1516–1520).
110. Weinstein, S., & Ebert, P. (1971). Data transmission by frequency-division multiplexing using
the discrete fourier transform. IEEE Transactions on Communication Technology, 19, 628–
634.
111. Falconer, D., Adachi, F., & Gudmundson, B. (1995). Time division multiple access methods
for wireless personal communications. IEEE Communications Magazine, 33, 50–57.
112. Li, W., Li, W., & Pan, Y. (2006). Resource Allocation in Next Generation Wireless Networks.
Wireless Networks and Mobile Computing. New York: Nova Science Publishers.
113. Working Group on Wireless Regional Area Networks, IEEE std. 802.22 (2011). http://www.
ieee802.org/22/.
114. Hu, W., Willkomm, D., Abusubaih, M., Gross, J., Vlantis, G., Gerla, M., et al. (2007). Cog-
nitive radios for dynamic spectrum access - dynamic frequency hopping communities for
efficient IEEE 802.22 operation. IEEE Communications Magazine, 45, 80–87.
115. Han, Z., & Liu, K. (2008). Resource Allocation for Wireless Networks: Basics, Techniques,
and Applications. Cambridge: Cambridge University Press.
116. Ge, Y., Sun, J., Shao, S., Yang, L., & Zhu, H. (2010). An improved spectrum allocation
algorithm based on proportional fairness in cognitive radio networks. In 2010 12th IEEE
International Conference onCommunication Technology (ICCT) (pp. 742–745).
117. Sengupta, S., Brahma, S., Chatterjee, M., & Shankar, N. S. (2013). Self-coexistence among
interference-aware IEEE 802.22 networks with enhanced air-interface. Pervasive and Mobile
Computing, 9(4), 454–471.
118. Sengupta, S., Chandramouli, R. Brahma, S. & Chatterjee, M. (2008). A game theoretic frame-
work for distributed self-coexistence among IEEE 802.22 networks. In IEEE Global Telecom-
munications Conference, 2008. IEEE GLOBECOM 2008 (pp. 1–6).
119. Coffman, E., Robert, P., Simatos, F., Tarumi, S., & Zussman, G. (2010). Channel fragmen-
tation in dynamic spectrum access systems: A theoretical study. Proceedings of the ACM
SIGMETRICS International Conference on Measurement and Modeling of Computer Sys-
tems, SIGMETRICS ’10 (pp. 333–344). New York: ACM.
120. Coffman, E., Robert, P., Simatos, F., Tarumi, S., & Zussman, G. (2012). A performance
analysis of channel fragmentation in dynamic spectrum access systems. Queueing Systems,
71(3), 293–320.
121. Karol, M., Haas, Z., Woodworth, C., & Gitlin, R. (1997). Time-frequency-code slicing: effi-
ciently allocating the communications spectrum to multirate users. IEEE Transactions on
Chapter 2
Cognitive Radio Network- A Review
As discussed in Chap. 1, the quality of service and energy efficiency of a cognitive

radio system depends upon various parameters, including spectrum selection, media
access scheme and spectrum sensing order. In this chapter, we provide a review
on technologies which facilitates/improves the parameters responsible for QoS and
energy management. At the end of this chapter, we will discuss about different
cognitive radio platforms and their evolution.
2.1 Spectrum Management
Channel selection (spectrum management) forms an integral part of quality of service

framework for cognitive radio network. Since different channels have different local
policy restriction, their usage offer different grade of services depending upon the
operating frequency [1]. Channel usage behavior of incumbents can be characterized
as deterministic or stochastic, leading a secondary user to utilize different etiquettes
for efficient spectrum utilization.
2.1.1 Ant Colony Optimization Based Spectrum Management
A channel selection technique is discussed in [2] which enables each secondary user
(SU) to detect appropriate channels over which it can communicate without any
coordination with the other SUs. The overall aim is to increase the spectrum utiliza-
tion in the system. The authors discussed a biologically-inspired Spectrum Sharing
(BIOSS) algorithm based on the adaptive task allocation model of insect colonies.
Here the problem of channel allocation is reduced to ant colony optimization. Table
2.1 shows the mapping between CRN and insect colony.
An ant performs efficiently if the suited tasks are allocated to them individually
and are allowed to simultaneously complete the task with other ants. Thus the col-
40 2 Cognitive Radio Network- A Review
Table 2.1 A mapping Insect colony Cognitive radio network

between ant colony and
cognitive radio Insect Cognitive radio
Task Available channel
Task associated stimuli (s) Permissible power to channel
(P j )
Response threshold (θ) Required transmission power
( pi j )
Table 2.2 A mapping Condition Events

between ant colony and
cognitive radio P j pi j Probability of cognitive radio
i to select channel j increases
P j pi j Probability of cognitive radio
i to select channel j
decreases
For constant pi j , if P j Probability of cognitive radio
increases i to select channel j increases
For constant P j , if pi j Probability of cognitive radio
increases i to select channel j
decreases
laborative behavior of the ants optimizes the overall job of ant colony. This division
if labour is based on the response threshold θ of each ant, which is the probability
of reacting to a task associated stimuli(s); whereas task stimuli(s) is the intensity of
activator associated with the task. The task is performed iff s θ.
The probability of performing a task is given by:
sn
Tθ (s) = (2.1)
s n + θn
Here, n > 1 determines steepness of the threshold.

Each cognitive radio i in the system calculates the channel selection probability
for every available channel c as:
csp P jn
Ti j = (2.2)
P jn + α pinj + β L inj
where, L i j : learning factor to enable CR i to learn or forget channel j and α, β:

positive constants.
The channels are selected on the basis of conditions as shown in Table 2.2.
The spectrum selection is done by Algorithm 2.1 which executes over each cog-
nitive radio node to provide spectrum selection based on QoS requirements.
2.1 Spectrum Management 41
Algorithm 2.1: Algorithm for load balancing in a given category of channels

1 Determine available spectrum bands;
2 Estimate P j values for every channel j;
3 Set learning factor (L i j ) value to be same for each channel and learning coefficients (ξ0 , ξ1 )
to appropriate values [2].
4 Computes T csp for all available channels.
5 Selects channels with max T csp .
6 if Selected channel meets QoS requirements then
7 L i j = L i j − ξ0 if selected channel meets QoS requirements.
8 else
9 L i j = L i j + ξ1 if selected channel does not meet QoS requirements.
10 end
Some of the observation worth to note is that, as the interference increase, P j for a
channel decrease. The channel with small P j due to the interference can be preferred
by the cognitive radio node which needs low transmission power. This will enable the
cognitive radio node to control the interference in the channels. Also, the spectrum
sharing becomes more efficient as the power demand of CR increases which inturn
would prefers the channel with more transmission power.
2.1.2 Non-linear Optimization Based Spectrum Management
In [3], the authors have considered a spectrum assignment problem for cognitive radio
sensor network via a Mixed Integer Non-linear Programming problem formulation.
The spectrum assignment problem is considered under coverage, interference, min-
imum data rate and power budget constraints. The problem is transformed to Binary
Linear Programming problem and further reduced to random, greedy and two-stage
(decoupled) algorithms.
The network model comprises of a set of N cognitive radio-enabled wireless
sensor nodes distributed within a set C of cells and coexisting with PUs operating in
an up-link network on the same band, and transmitting on each channel with power
PPU . Both systems operate on a band comprising a set of K channels, each with
bandwidth Bk , ∀k ∈ K . Only one node is active to report the measured data while
others are in sleep mode.
|h kn,SU |2 is the channel gain between node n and the cognitive radio base station
receiver on channel k. |h kPU |2 is the the channel gain between the primary user
transmitter and the secondary user base station receiver on channel k. No is the
power spectral density of a single sided additive white Gaussian noise. S is the set
of active nodes.
Each of the active sensors from different cells is assigned channels in order to
minimize transmit power with the following constraints:
(i) Interference at primary network receivers is less than given threshold. That is,
Pnk |h kn,PU |2 ≤ Ithk , ∀k ∈ K , ∀n ∈ N , where Pnk is the transmission power of
node n on channel k, |h kn,PU |2 is the channel gain between node n and the primary
user base station receiver on channel k, and Ithk is the interference threshold at
PU on channel k.
(ii) Mutual exclusion of channels among cognitive N radio sensor nodes (A channel
k is used by at most one node). That is, n=1 Ynk ≤ 1, ∀k ∈ K , where Ynk is a
binary variable identifying whether a node n is assigned channel k.K
(iii) Transmission power is limited by a power budget. That is, k=1 Ynk Pnk ≤
Pn , ∀n ∈ N , where Pn is the maximum allowed power of node n.
max max
(iv) Secondary network must achieve a minimum data rate (for QoS). That is,

K
Ynk .Bk .log2 (1 + Pnk ∗ γnk ) ≥ Rth (2.3)
k=1
where γnk is the channel to interference noise ratio of node n on channel k, and
Rth is the minimum rate required required at the cognitive radio base station
receiver.
(v) A coverage constraint where the geographical area is divided into multiple
cells, each covers a certain
N event, and each cell must be covered by one active
sensor node. That is, n=1 Acn Yn = 1, ∀c ∈ C, where Acn is a binary variable
identifying
K whether a cognitive radio sensor node n is assigned a channel k, and
Yn = k=1 Ynk , ∀n ∈ N , indicates whether a node is active or not.
The problem of choosing active sensors and assigning channels to these sensors can
be formulated as:

N
K
min Ynk Pnk
Ynk ,Pnk
n=1 k=1
subject to: Pn |h n,PU |2 ≤ Ithk , ∀k

k k
∈ K , ∀n ∈ N ,

N
Ynk ≤ 1, ∀k ∈ K ,
n=1

K
(2.4)
Ynk Pnk ≤ Pnmax , ∀n ∈ N ,
k=1

K
Ynk .Bk .log2 (1 + Pnk ∗ γnk ) ≥ Rth , ∀n ∈ N ,
k=1

N
Acn Yn = 1, ∀c ∈ C
n=1
where,
1, if sensor n cover area c,
Acn =
0, otherwise

1, if sensor n is assigned channel k,
Ynk =
0, otherwise
|h kn,SU |2
γnk =
No ∗ B + PPU ∗ |h kPU |2
The problem in Eq. 2.4 being NP-hard cannot be solved in polynomial time. Since
cognitive radio sensor nodes are resource constrained, a realistic assumption regard-
ing single transceiver on each node can be made. This would reduce the above
multi-integer linear program to binary linear program as shown in Eq. 2.5.

N
K
min Ynk Pnk
Ynk ,Pnk
n=1 k=1
subject to: Pnk |h kn,PU |2 ≤ Ithk , ∀k ∈ K , ∀n ∈ N ,

N
Ynk ≤ 1, ∀k ∈ K ,
n=1
(2.5)
Pnk ≤ Pnmax , ∀k ∈ K , ∀n ∈ N ,
Bk .log2 (1 + Pnk ∗ γnk ) ≥ Rth , ∀k ∈ K , ∀n ∈ N ,

N
Acn Yn = 1, ∀c ∈ C
n=1
Algorithm 2.2: Algorithm for calculation of Pnk

1 Calculate γnk , |h kn,PU |2 , ∀k ∈ K , ∀n ∈ N ;
2 Calculate Pnk :
3 for n = 1 to N do
4 for k = 1 to K do
Rth /Bk
5 Pnk = Pnk |min = 2 γ k −1 ;
n
k
Ith
6 if Pnk |min > min then
|h n,PU |
k 2
7 Pnk ⇐ ∞;
8 end
9 end
10 end
Using Algorithm 2.2, the transmission power of each node n over all channels avail-
able in the network can be calculated. This can be utilized to generate a matrix Q,
which provides transmission power of cognitive radio nodes in each cell. This can
be given as Q = Pnkc , ∀n ∈ N , ∀c ∈ C, ∀k ∈ K and can be calculated as given in
[3]. The minimization problem given in Eq. 2.5 is reduced to a simpler minimization
problem (Eq. 2.6) by utilizing the matrix Q.
For given Pnk , ∀n, ∀k

N
K
min Ynk Pnk
Ynk
n=1 k=1

N
subject to: Ynk ≤ 1, ∀k ∈ K (2.6)
n=1

N
Acn ∗ Yn = 1, ∀c ∈ C
n=1
The discussed spectrum selection scheme utilizes following algorithms to solve the
optimization problem in Eq. 2.6.
2.1.2.1 Minimum Power Algorithm
The aim of the Minimum power algorithm is to jointly optimize the selection of
nodes and channels in such a way that the total transmission power is minimized.
The transmission powers Pnk from Algorithm 2.2 is utilized in Eq. 2.6 as shown in
Algorithm 2.3.
Algorithm 2.3: Minimum Power Algorithm

1 Calculate γnk , |h kn,PU |2 , ∀k ∈ K , ∀n ∈ N
2 Calculate Pnk from Algorithm 2.2;
3 Calculate matrix Q from Pnk ;
4 Node/Channel Selection:
5 Solve Eq. 2.6 with the values of Q;
Assuming total number of bits in a binary representation for each of the optimal
variables to be 1, the overall complexity of this algorithm is O((N K )3.5 ).
2.1.2.2 Random Algorithm
The random algorithm choses a random node out of each cell and then assigns chan-
nels to it randomly. This can be achieved by the method discussed in
Algorithm 2.4.
Algorithm 2.4: Random Algorithm

4 Node Selection:
5 for c = 1 to C do
6 S(c) ⇐ select a random node;
7 end
8 Channel Selection:
9 for c = 1 to C do
10 S(s)k ⇐ select a random channel k ∈ K ;
11 K ⇐ K − k;
12 end
The complexity of this algorithm is given as O(1) for random selection of

node/channel.
2.1.2.3 Greedy Algorithm
The greedy algorithm performs search over the matrix of Pnk obtained from Algorithm
2.2 to find the node/channel pairs that has minimum transmission power. This can
be achieved via the following Algorithm 2.5.
Algorithm 2.5: Greedy Algorithm

4 Node/Channel Selection:
5 for c = 1 to C do
6 S(c)k = min(Pnk ), ∀n ∈ N ∀k ∈ K ;
7 Q ⇐ Q − Q(:, k);
8 Q ⇐ Q − Q(c, :);
9 K ⇐ K − k;
10 end
The total complexity of this greedy algorithm is of O(C(N K )2 ).
2.1.2.4 Two-Stage (Decoupled) Algorithm
In this method, the problem of spectrum allocation is broken into two sub-problems:
node-selection and channel-assignment. In node-selection problem, life of node is
extended by selecting nodes with higher residual energy as cells, while channel-
assignment problem is solved as a standard linear program.
Algorithm 2.6: Two-Stage Algorithm

4 Node Selection:
5 for c = 1 to C do
6 S(c) = max(n Er es ), ∀n ∈ Nc ;
7 end
8 Channel Selection:
9 Solve Eq. 2.6 with Q s as input.
The overall complexity of this two-stage scheme is O((C N 2 ) + (C K )3.5 ).
2.1.3 Game Theory Based Spectrum Management
A solution to address the problem of efficient spectrum pricing in cognitive radio

network is given in [4]. The inefficiency of the Nash equilibrium (equilibrium pricing
adopted by primary service providers) is discussed, in the sense that the profit of the
primary service providers is not maximized. Here an alternate solution to this problem
is discussed.
The problem is modelled as oligopoly market, in which a small number of firms
compete with each other non-cooperatively and independently to achieve their objec-
tives while dominating the overall market. The objective is determined in terms of
controlling the price of products/quantity, and such actions are influenced by the
decisions of other firms too. To achieve an equilibrium pricing scheme Bertrand
game model may be utilized. Similarly, pricing for primary users’ spectrum can also
be calculated in a similar way, which are willing to share with secondary users.
A wireless system with total number of primary services N is assumed in which
all primary services operate over different frequency spectrum Fi , ∀i ∈ N . Each
of the primary service here wants to sell a portion of the available spectrum Fi to
secondary users for price pi as shown in Fig. 2.1.
An adaptive modulation is utilized to dynamically adjust the transmission rate.
The spectral efficiency k of transmission by a secondary user is given by Eq. 2.7.
1.5
k = log2 (1 + K γ), where K = (2.7)
0.2
ln B E R tar
where γ is the SNR at the receiver and B E R tar is the target bit-error-rate.
Primary Primary Primary

Service 1 Service 2 Service n
Spectrum F1 Spectrum F2 Spectrum Fn
Price P2
Price Pn
Price P1
Secondary Service
User 1 User 2 User 3 User 4
Fig. 2.1 Spectrum sharing model
Utility of the shared frequency spectrum is used to calculate the spectrum demand
of the secondary users. A quadratic utility function is used to quantify the demand
or utility of the spectrum.
⎛ ⎞

N
1 ⎝ 2
N N
U (b) = bi ki(s) − bi + 2ν bi b j ⎠ − pi bi (2.8)
i=1
2 i=1 i= j i=1
where b is the set consisting of the size of shared spectrum from all the primary
services, i.e., b = b1 , . . . , b N ; pi is the price offered by primary service i, ki(s)
denotes the spectral efficiency of wireless transmission by a secondary user using
primary users’ frequency spectrum Fi . ν is the spectrum substitutability factor ν ∈
[−1.0, 1.0].
The quadratic utility function is used because of the following reasons:
(i) Quadratic function being concave, is able to represent the saturation of user
satisfaction as the transmission rate increases.
(ii) A linear bandwidth demand function can be derived via differentiating the
quadratic utility function. This eases the subsequent analysis for stability con-
dition.
(iii) The spectrum substitutability factor ν, control the switching of secondary users
onto its channels. If ν = 1, a secondary user can switch freely among operating
frequency spectra; while if ν = 0, a secondary user can not switch freely.
The demand function for spectrum Fi is given by differentiating U (b) with respect
to bi as shown in Eq. 2.9.
∂U (b)
= 0 = ki(s) − bi − v b j − pi (2.9)
∂bi i= j
The spectrum demand function is obtained from Eq. 2.9 by substituting the value of
the prices of the primary service as shown in Eq. 2.10.

(ki(s) − pi )(v(N − 2) + 1) − v (s)
i= j (k j − pj)
Di (p) = (2.10)
(1 − v)(v(N − 1) + 1)
Price charged by the primary users can also be used to calculate the spectrum demand
of the secondary users. The cost of the primary service in sharing the spectrum is
computed from the degradation of the QoS for the primary service. The cost is only
incurred if the required bandwidths for the primary connections cannot be provided
due to bandwidth sharing with secondary services. For primary service i, the revenue
function Ri and the cost function Ci , can be defined as follows:
Ri = c1 Mi
2
r eq ( p) Wi − bi (2.11)
Ci (bi ) = c2 Mi Bi − ki
Mi
Here c1 and c2 denote the constant weight for revenue and cost functions. Mi is the
r eq
number of primary connections, Bi is the bandwidth requirement and Wi is the
size of spectrum of the the primary service i.
In the above given oligopoly market model, Bertrand game model can be utilized
to formulate the game. Primary services act as the players, modifying the price per
unit of the spectrum ( pi ) act as the strategy, while the profit of sharing the spectrum
with secondary services act as the payoff. Therefore the profit of each primary service
can be calculated as shown in Eq. 2.12.
Pi (p) = bi pi + Ri − Ci (bi ) (2.12)
where p denotes the set of prices offered by all players in the game, that is p =
{ p1 , p2 , p3 , . . . , p N }.
The best response function (best strategy of one player for given strategies of
other players) of primary service i, given a set of prices offered by the other primary
services p−i is defined as follows:
Bi (p−i ) = ar g max Pi (p−i ∪ { pi }) (2.13)

pi
An equilibrium is reached iff

pi∗ = Bi (p∗−i ), ∀i (2.14)
where p∗ = { p1∗ , p2∗ , . . . , p ∗N } is the set of prices when Nash equilibrium will be
reached. p−i is the set of best responses for player j such that j = i.
From Eqs. 2.12, 2.10 and 2.11, the profit function can be derived as shown in
Eq. 2.15.
2
r eq ( p) Wi − Di (p)
Pi (p) = pi Di (p) + c1 Mi − c2 Mi Bi − ki (2.15)
Mi
The demand function in Eq. 2.10 can be expressed as: Di (p) = D1 (p−i ) − D2 pi .
Where D1 (p−i ) and D2 are given by Eq. 2.16.

ki(s) (v(N − 2) + 1) − v (s)
i= j (k j − pj)
D1 (p−i ) =
(1 − v)(v(N − 1) + 1) (2.16)
(v(N − 2) + 1)
D2 =
(1 − v)(v(N − 1) + 1)
For a Nash equilibrium to occur,
∂ Pi ( p)
= 0, (2.17)
δ pi
From Eq. 2.15, 2.16 and 2.17, the Linear Eq. 2.18 can be obtained.

( p) r eq ( p) Wi − (D1 (p−i ) − D2 pi )
0 = 2c2 ki D2 Bi − ki + D1 (p−i ) − 2D2 pi (2.18)
Mi
For given parameters, the Eq. 2.18 provides the solution pi∗ which provides Nash
equilibrium. From this the size of the shared spectrum can be calculated from Eq. 2.10
(i.e. Di (p∗ )).
In reality, primary services may not be able to observe the profit gained by other
primary services. In such scenario, for a Nash equilibrium to occur, a distributed
form of price adjustment is required. Let the price offered by primary service i at
iteration t be given as pi [t]. If the strategies of the primary services in the previous
iteration are observable by each other, then pi [t] is given as shown in Eq. 2.19.
pi [t + 1] = Bi (p−i [t]), ∀i (2.19)
If the strategies used by other primary services are not known, then the relationship
between the strategies in the current and the future iteration is to maximize the profit,
which can be expressed as follows:

∂ Pi ( p)
pi [t + 1] = pi [t] + αi (2.20)
∂ pi
where αi is the learning rate.

The marginal profit of a primary service can be observed by the marginal spectrum
demand for small variation in price ε as shown in Eqs. 2.21 and 2.22.
∂ Pi (p) Pi (p−i [t] ∪ { pi [t] + ε}) − Pi (p−i [t] ∪ { pi [t] − ε})

≈ (2.21)
∂ pi 2ε
Pi (p−i [t] ∪ { pi [t] ± ε}) = pi Di (p−i [t] ∪ { pi [t] ± ε}) + c1 Mi

2
r eq ( p) Wi − Di (p−i [t] ∪ { pi [t] ± ε})
− c2 Mi Bi − ki
Mi
(2.22)
To check if the above developed dynamic game will reach Nash equilibrium at the
steady state, stability of the game is evaluated. Stability can be evaluated by utilizing
the eigenvalues of the Jacobian matrix of a self-mapping function. The self mapping
function is stable if the eigenvalues λi are inside the complex unit circle. Here the
self mapping function utilized is from Eq. 2.20. Therefore the Jacobian matrix is
defined as: ⎡ ∂ p [t+1] ∂ p [t+1] ⎤
1
∂ p1 [t]
1
∂ p2 [t]
· · · ∂∂p1p[t+1]
N [t]
⎢ ⎥
⎢ ∂ p2 [t+1] ∂ p2 [t+1] · · · ∂ p2 [t+1] ⎥
⎢ ∂ p1 [t] ∂ p2 [t] ∂ p N [t] ⎥
J =⎢ . .. .. ⎥
⎢ . ⎥
⎣ . . . ⎦
∂ p N [t+1] ∂ p N [t+1]
∂ p1 [t] ∂ p2 [t]
· · · ∂ ∂p Np[t+1]
N [t]
It can be observed that, if the strategies of other primary services are observable,
then there is no control parameter to stabilize the price adaptation. If the strategies of
other primary services are not observable, then the control parameter depends upon
the learning rate.
For a repeated game, a trigger strategy allows a primary service to maintain
collusion as long as other services agree to do so. In the case that the collusion is
maintained forever, the long-term profit of primary service i can be expressed as
follows:
1
Pi0 + δi Pi0 + δi2 Pi0 + · · · = P0
1 − δi i
where Pi0 is the profit due to the optimal price, and δi is the weight (0 ≤ δi ≤ 1)
which multiplies the profit in next stage.
For a game where one primary service deviates from the optimal price, it will
experience the profit at the Nash equilibrium during the rest of the stages. The long
term profit in this case can be expressed as:
δi
Pid + δi Pin + δi2 Pin + · · · = Pid + Pn
1 − δi i
where Pid and Pin are the profit due to price deviation and profit due to price at the
Nash equilibrium respectively.
Collusion will be maintained if:
1 δi
Pi0 ≥ Pid + Pn
1 − δi 1 − δi i
2.1.4 Learning Automata Based Spectrum Selection
A spectrum selection scheme is discussed in [5], which minimizes the overall switch-
ing of the spectrum done by secondary users. This is done so as to mitigate the problem
of aggravated delay and deteriorated packet-loss ratio which occurs due to constant
channel switching of sporadically varying channel in cognitive radio environment.
This is further complicated by the governing guarantee on maximum interference
that can be caused to primary users by secondary users. Therefore a secondary user
has to vacate a channel within a given time threshold whenever it detects the presence
of primary users.
To minimize channel switching, the discussed stochastic channel selection algo-
rithm utilizes learning automata which maximizes the number of successful trans-
missions. The algorithm asymptotically converges to an optimal channel which has
higher availability in comparison to other channels.
In the given system, a secondary user can tolerate interference up to a threshold
value; once the threshold value is reached, the secondary user has to leave the system.
Whenever the channel is selected the secondary user transmits over consequent time
slots, until the channel ceases to be free, Thus, a transmission is successful, if:
(i) The primary user is absent throughout the time slot.
(ii) Any of the neighbouring secondary user (the secondary users whose actions
may potentially cause a transmission failure of the given secondary user) isn’t
operating at the same frequency.
The cognitive radio system is modelled as learning automata based control system,
where the channel availability is modelled as random environment while the finite
actions are modelled as the spectrum decisions. Each of these deterministic finite
action will induce a random output from the random environment, which could be
either favorable or unfavorable (i.e. whether selected channel has high availability or
not). A learning automata algorithm can be utilized over this output to provide a con-
trol over the decision making of the finite actions. The learning automata algorithm
essentially utilizes past results to make decisions about the next action.
During initialization of the secondary network, a secondary user selects any avail-
able channel at the beginning of each time slot for transmission and the output is
denoted as β (β = 0 for successful and β = 1 for unsuccessful transmission). The
system comprises of M primary users operating over C channels on which secondary
users coexist in a hierarchical way. The system is viewed from the perspective of a
secondary user denoted by k. The starting time of each slot is denoted by Ti . Each
secondary user maintains a probability vector denoted by Pk = [ p1 , . . . , pm ] where
pi is probability of choosing channel i. Each secondary users also maintains an esti-
mation vector denoted by Dk = [d1 , . . . , dm ], where di is the estimated possibility of
successful transmission by selecting channel i. H (n) denotes the number of channels
having higher values in the estimation vector D than the current selected channel. R
and W are the resolution parameter and the initialization parameter respectively. δ
represents the step size of adjusting the probability vector where, δ = 1/R. Si (n) is
the number of slots where the transmissions with selected channel i are successful,
up to Tn . Ci (n) is the number of slots where the channel i are selected to transmit
packets, up to Tn .
The Algorithm 2.7 shows the learning automata based channel selection technique
which executes over each node k with B as a predefined convergence threshold.
This channel selection technique is ε-optimal for any cognitive radio networks
which is stationary. Therefore for any arbitrarily small ε > 0 and γ > 0, there exists
a n satisfying:
Pr {|1 − pz (n)| < ε} > 1 − γ, ∀n > n
where z is the index of the optimal channel in terms of probability of successful

transmissions.
Algorithm 2.7: Algorithm for Stochastic Channel Selection

1 Initialize:
2 P(0) = [ p1 , . . . , p M ] where pi = 1/M for all 1 ≤ i ≤ M;
3 do
4 Select available channel from P(0) ;
5 Evaluate and record β ;
6 while Until each channel is selected W times;
7 Update Si (0) and Ci (0) according to value of βs for each channel i ;
8 Initializes D(0) = [d1 , . . . , d M ] where di = Si (0)/Ci (0) ;
9 do
10 Select an available transmission channel i according to the probability vector P(n) ;
11 Update the probability vector P(n) as:

12 p j (n + 1) = min p j (n) + H δ(n) , 1 ∀ j if d j (n) > di (n)
δ

13 p j (n + 1) = max p j (n) − M−H (n) , 0 ∀ j if d j (n) < di (n)

14 pi (n + 1) = 1 − j=i p j (n + 1)
15 After Tn , the secondary user adjusts the value of D as:
16 Si (n + 1) = Si (n) + (1 − β)
17 Ci (n + 1) = Ci (n) + 1
18 di (n + 1) = CSii (n+1)
(n+1)
19 while (max(P(n)) > B);
2.1.5 Spectrum Selection in Varying Channel Bandwidth

Environment
In [6], a scheme for cognitive radio spectrum hand-off is discussed where spectrum
switching is done between channels of varying bandwidth. The concept of delay
bandwidth product (DBP) is utilized to prioritize the channels with varying band-
width. Here the delay bandwidth product is defined as the difference between the
maximum tolerable delay threshold of the secondary user and the average occupation
time of the primary user over a given channel. Based on this a secondary user can
make spectrum selection to reduce power, increase range, increase throughput, load
balance and guarantee its QoS requirement.
Theoretically delay bandwidth product refers to product of a data link’s capacity
and its round-trip delay time. Intuitively, this indicates the maximum amount of
data that can exist in a network at any given time. Here the delay is considered as
the difference between maximum tolerable delay of a secondary user and average
occupation time of a primary user. As shown in Fig. 2.2, the total delay time (Di ) is
the elapsed time until the secondary user gets the opportunity to transmit again. The
total delay time is dependent on sensing time W j , handoff time t0 and the transmission
time of secondary user Tk , where k is the current selected channel and j is candidate
sensed channel. A secondary user can transmit again on the same selected channel
or it can switch to another channel depending upon the DBP index.
The system comprises of N variable bandwidth channels each with bandwidth
Bi , ∀i ∈ (1, 2, . . . , N ). The current channel is defined as k while the candidate
channel is defined as j. The total delay time Di of secondary user i can be expressed
as shown in Eq. 2.23.
Fig. 2.2 Channel selection by secondary user


t0 + W j , if 1 ≤ j ≤ N , j = k;
Di = (2.23)
Tk , j =k
If the sensing time is less, the transmission time increases. Also for a higher band-
width channel, there is an increase in throughput. Therefore an optimization is
required between the bandwidth of the channel and the effective delay required
by the channel itself. The values of Tk and W j are dependent on what kind of traffic
model is utilized in the system.
Two different primary user traffic models are considered for channel availability:
Pareto distribution model and Two-state Markov chain model.
1. Pareto distribution model:
The probability density function (f(x)) and cumulative distribution function (F(x))
are described as shown in Eq. 2.24.
λK λ
f (x) = ,x ≥ K
x λ+1
(2.24)
1 − (K /x)λ , x > 0
F(x) =
0, otherwise
where λ is the scale parameter and K is the shape parameter, λ > 0, K > 0.
2. Two-state Markov chain:
A two-state Markov chain model is utilized to model the primary user traffic behav-
ior. Here PI and PB represent channel transition probability to being available and
busy respectively. 1 − PI and 1 − PB represent transition probability from state of
availability to unavailability and vice versa.
The steady state probability of channel i to be busy by sending primary user’s
traffic is given by:
B (i) = p B (1 − p B )
The average sensing time of a SU Wi is given by:

TT
hr eshold
Wi = L P(Ts = L) (2.25)
L=1
where, TT hr eshold is the number of slot of secondary user’s maximum channel sensing
tolerance and Ts is the number of slots spend by secondary user for sensing the
available channel.
A control parameter Ci is used to increase or decrease DBP index according to the
channel conditions and the channel bandwidth ratio in reference to other channels’
bandwidth. It is also used to track the difference between the data rates of different
channels. Ci value starts from one for all channels and updates as primary user
appears on channel i. The successful transmission probability Pi is calculated as the
percentage of successful transmitted slots to the total number of transmitted slots on

channel i. The short term updates of data rate (Ri) of channel i can be given as:

Ri (1 − α), if Pi ≤ 0.9;
Ri =
Ri (1 + α), if Pi > 0.9;
where i varies from 1 to N . α is the rate smoothing parameter and its value is
0.001. Ri increases with increase in successful transmission probability Pi > 0.9.
Similarly, Ri decreases for transmission probability Pi ≤ 0.9 which indicates bad
channel condition. The long term updating of Ci is done every 50 slots according to
the following rule:
⎧
⎪ 1 N Rj
⎪
⎨ i
C − C, if Ri
R∗
− N j=1 R ∗ < −ε
Ci =
⎪
⎪ N R j
⎩Ci + C, if R∗i − 1 j=1 R ∗ > ε
R N
where R ∗ is the target rate, i is the current channel, j is the candidate channel, ε is
threshold value, and C is a fixed step parameter.
If Tmax is the maximum delay allowed by secondary user and Tavg is the prede-
termined average time, then for i varying from 1 to N , the priority index ηi can be
expressed as:

(Tmax − Di )Bi , Di ≤ Tavg
ηi =
Ci (Tmax − Di )Bi , otherwise
ηi represents the maximum capacity of channel i. This priority index represents the
delay bandwidth product based scheme, therefore priority index increases as the DBP
increases.
The channel selection when primary user appears on the current selected channel
is given as:
Channeli = max{ηi }
This selected Channeli becomes the candidate channel k.

The performance of this DBP-based scheme is calculated to determine whether
it meets the required service and reliability objectives. The effective data rate Re f f
can be calculate from Eq. 2.26.
N
tsi Bi
Re f f = i=1 (2.26)
ttotal
where tsi is the successful transmission slot, Bi is the average bandwidth of channel
i, while ttotal is the total transmission time and can be calculated from Eq. 2.27

N
N
ttotal = tsi + Di (2.27)
i=1 i=1
where tsi is successful transmission slot and Di is total delay time of secondary user i.
The total effective data rate will be maximized as long as the cognitive radio node
stays over the channel with the highest DBP index.
2.2 Media Access Control
As discussed in Sect. 1.7, a media access scheme (MAC) is required among multiple
secondary users who are sharing a channel so as to mitigate access collision. Media
access control protocols can be broadly classified into two categories: random and
time-slotted media access schemes, each having its pros and cons. Recently a new
class of access scheme, known as hybrid schemes has been developed with better
performance than that of the random and time-slotted access schemes.
2.2.1 QoS Aware Media Access Schemes
A QoS aware MAC protocol for cognitive radio with real time delay sensitive mul-
timedia applications is discussed in [7]. The discussed scheme determine the set of
channels for sensing and transmission of data which will satisfy the QoS requirements
of the cognitive radio user. The authors also implements a priority based spectrum
access scheme at the secondary user level to provide the QoS guarantee.
The protocol is utilized in a single hop cognitive radio network environment
where all the secondary users can communicate with each other. The cognitive radio
networks comprises of N c data channel and 1 control channel, which is selected
in the unlicensed UWB (ultra wide band) spectrum. A control channel is used to
to synchronise and agree on a data channel for transmission and then tune to that
data channel for communication. Each secondary user is equipped with a single
transceiver and therefore can only sense one channel at a time. Each channel is
modelled as an ON/OFF model. Secondary users keeps track of the slowly varying
characteristics and usage patterns of the primary users.
As discussed in Sect. 1.5.1, the secondary user senses the channel for an arbitrary
time, the arbitrary sensing period (ASP) which includes a basic sensing time to ensure
sensing accuracy and a random slot selected from window [0, SWi ]. This is done
to mitigate access collisions among all secondary users who are accessing the same
channel. If a primary users shows up on the same channel utilized by the cognitive
radio during the data transmission phase, the transmission fails and the secondary
2.2 Media Access Control 57
user has to vacate the channel within a given time threshold. If the secondary user
wants to continue transmission of data, the secondary user has to move to some other
channel for data communication.
The transmission time of the current frame of user i is given as ti , then the
probability that the current frame is transmitted over the channel k is
βk
Pik = Fk (t ASi + ti ) (2.28)
βk + αk
where 1/αk and 1/βk are the mean idle and busy periods of channel k respectively,
t ASi is the arbitrary sensing period selected by user i, ti is the transmission time for
the current frame, and Fk (x) is the cumulative distribution function of the usage of
channel k by the primary users. It is assumed that no more than one primary user
will transmit at the same time on the given channel. Therefore, the transmission by
a cognitive radio user is successful if no primary user appears during the t ASi + ti
interval.
Once the transmission probabilities are computed, a channel sensing sequence can
be designed. Two different channel sensing policies have been discussed: greedy
and ascending. In the greedy technique, all the channels are sorted in the descending
order and the secondary users selects channel in that order. This guarantees maxi-
mum probability of successful transmission for each secondary user which will result
in increased selection of this channel by secondary users. This increase in selection
will ultimately lead to increase in contention which will significantly reduce the
throughput. The ascending approach allows different secondary users to select vari-
ous channels in the order of the available channel probability. For a given available
channel probability of channel k for user i (Pik ), the expected transmission time over
t +t
channel k is given as E[Tik ] = ASPi k i . The secondary user selects the channels which
i
will satisfy the condition E[Tik ] < τi , where τi is the maximum tolerable one hop
delay which a secondary user can guarantee for QoS provisioning. Because of this,
the delay performance can be guaranteed by the ascending policy.
To improve the QoS, service differentiation is applied in terms of size of sensing
window. A cognitive radio user differentiate the traffic based on different sensing
window size. A higher priority application with higher QoS constraints will have a
shorter sensing period so that they have an higher chance of accessing the data channel
when the opportunity appears. (i.e., SWvoice < SWvideo < SWdata ). Multiple levels
of QoS provisioning can be defined by defining different sensing window sizes for
different traffic types as shown in Table 2.3.
For a secondary user, the transmission fails for the current time slot on channel k
if any of the following conditions are true:
(i) The channel is already occupied by a primary user. The probability of which
can be given as:
αk
k
Poccupied = (2.29)
αk + βk
Table 2.3 Sensing windows for different multimedia services

Type Strict priority Statistical priority No priority
Voice [0, 31] [0, 31] [0, 31]
Video [32, 63] [0, 63] [0, 31]
Data [64, 127] [0, 127] [0, 31]
(ii) A primary user occupies the channel in middle of channel sensing. The proba-
bility of which can be given as:
βk
k
PPU = · Fk (t ASi ) (2.30)
αk + βk
(iii) The channel gets busy due to another secondary user transmission. This will
depend upon the sensing window size. Here for a maximum sensing window
size of W and Nk secondary users contending on channel k, the probability that
the given secondary user wins the contention is given as:
W
1 W − j Nk
Pr (given secondary user wins contention) = (2.31)
j=1
W W
Since every secondary user is contending on the channel k, it implies that there
were no primary users’ detected on that channel. Therefore the probability that a
secondary user’s transmission fails due to another secondary user transmission
is given as:
⎛ ⎞
Nk
βk W
1 W − j
k
PSU = 1 − Fk (t ASi ) ⎝1 − ⎠ (2.32)
αk + βk j=1
W W
Considering all three cases, the probability that the transmission from a secondary
user is successful on channel k is given as:

k
Psuccess =1− k
Poccupied + k
PPU + k
PSU (2.33)
With this, the average time a secondary user occupies a channel k is given as:
E[T k ] = Psuccess
k
(t ASi + ti ) + (1 − Psuccess
k
)t ASi (2.34)
A transmission is successful only if no primary user shows up during the total sensing
and transmission time of the secondary user. The probability of which can be given
as:
W
βk 1 W − j Nk
PTk S = Fk (t ASi + ti ) (2.35)
αk + βk j=1 W W
The probability that a secondary user is able to succeed in transmission in r -th attempt
on the channel k is given as:
r −1

PT S (r ) = PTk S (1 − PTk S ) j (2.36)
j=1
The average transmission delay of a secondary user over a given channel k is given
as:
∞

E[T ] = E[T k ]PT S (r ) (2.37)
r =1
The discussed media access scheme along with channel sensing scheduling scheme
is QoS aware, which provides differentiated services for multimedia and voice. It
provides real time service guarantees for traffic with higher priorities. Since it is
utilized for single hop network, in multiple hop network, the protocol will suffer
from the problem of hidden terminal.
In [8], two media access scheme are discussed for cognitive radio networks which
provisions certain level of QoS to voice packets in the network. The two media access
scheme are based on contention and time-division.
The contention based medium access protocol is similar to various protocols that
have been already discussed. The channel time is partitioned into time slots, and a
time slot is further divided into four phases: sensing, contention, transmission and
ACK. The first phase (sensing phase), is used for all the secondary users to sense
the activity of the primary users. If the channel is sensed busy, no secondary user
contends for that slot. The second phase (contention phase), comprises of a number
of minislots. All the secondary users have the same contention window size. If the
channel is sensed idle in the sensing phase, a secondary user randomly chooses a
backoff timer from the contention window and continues to sense the channel in the
contention phase. If the channel is sensed to be continuously idle for a duration of the
backoff timer (in units of minislots) in the contention phase, the secondary user will
terminate during the contention phase and transmit its packet in the third phase of the
time slot (transmission phase); otherwise, it will quit the contention for the current
slot. Therefore, for each contention, the secondary user with the smallest backoff
timer will win and transmit its packet in the transmission phase of the time slot. If
more than one secondary user choose the same smallest backoff timer, it will result
in a collision. To determine whether a packet has successfully been transmitted, the
receiver sends an acknowledgment (ACK) at the ACK phase of each slot to the sender
upon a successful packet reception.
The time-division based media access depends on dividing the time into indexed
mini slots. Each mini slot is assigned to a secondary user in a deterministic way such
that the secondary user can only transmit if the channel is idle from minislot 0 to
i − 1. Therefore a secondary user has to sense the channel from 0 to i − 1 and can
only transmit if the channel is found idle. This scheme also means that the CR users
with higher indexed mini slots will have a lower chance of getting a transmission
opportunity and hence the indexing is rotated periodically. If the number of users in
the CR network are always same, mini slot assignment can be done at initialization.
If instead the number of users change dynamically, the protocol can utilize control
packets for joining and leaving the network.
A 3-dimensional Markov chain is utilized to do numerical analysis where each
state of the Markov chain is represented by (n, t, s), where s = 0 is the idle state and
s = 1 is the busy state. When n > 0, t is the queuing delay experience by the queued
packet, and when n = 0, t is the time required for the arrival of the next voice packet
where n indicates the number of packets in the queue.
The probability of success for the contention based protocol in independent
channel-state model is derived as:
⎡ ⎤

N s −1
CW i
Ns − 1 i 1 CW − j ⎦
Ps = θ N p ρ (1 − ρ) Ns −i−1 ⎣ (2.38)
i=0
i j=1
C W CW
where θ N p is the probability that a channel is vacated by individual primary users and
ρ is the queue utilization of the secondary user. Ns , is the total number of secondary
users and C W is the contention window size of secondary users.
The probability of success for the contention-free (time-division) protocol in inde-
pendent channel-state model is derived as:
Ns
1
Ps = θ N p (1 − ρ)i−1 (2.39)
i=1
N s
The packet drop probability in independent channel-state model is calculated as:

π(n, t)(1 − Ps ) D−t
Pdr op = n>0 (2.40)
n>0 π(n, t)
where π(n, t) is the steady state probability of state (n, t). Then, the capacity for the
voice secondary users Ns∗ can be obtained by finding the maximum integer value of
Ns that satisfies Pdr op ≤ Pl .
The method ensures the packet loss if the number of users are within the allowed
limits along with the provisioning of certain level of QoS the the voice users. The
protocol has very high overhead due to maintaining indices and dynamically changing
indices in the time-division based media access scheme. The performance of the
network would suffer when nodes leave and rejoins the networks.
In [9], a media access control scheme for multi-hop cognitive radio network is
discussed. This MAC scheme aims at increasing the throughput in a distributed
cognitive radio environment while utilizing the power saving mode operation for
conserving energy.
The media access control utilizes CSMA/CA based channel access with distrib-
ution coordination function for distributed multi-channel access. The media access
scheme also utilizes a power saving mode (PSM) operation for conserving power.
The idea of PSM is to let the nodes enter a low powered dozing state if no com-
munication is required. As discussed in Sect. 1.7.1, time in the system is divided
into superframes, where each superframe is divided into AT I M window and Data
window. The AT I M window also comprises of a beacon interval. At the start of
each beacon interval, any node A which has to send a packet to a node B sends an
ad-hoc traffic indication message ( AT I M) to node B within this window. When B
receives the packet, it sends an ATIM-ACK frame and A and B stay awake for the
rest of communication duration. Nodes that did not send or receive a AT I M frame
enter a doze state till the next beacon period. The AT I M window can be represented
as contention window as discussed in Sect. 1.7.1 and shown in Fig. 1.9. Each of
the cognitive radio user utilizes a dedicated sensing hardware for sensing purposes.
Therefore a cognitive radio user has the ability to perform out-of-band sensing. We
will discuss about out-of-band sensing later in Sect. 3.4.2.
The system utilizes distributed collaborative sensing to detect primary user inter-
ference and to avoid the hidden node problem. The system comprises of N secondary
users and utilizes C + 1 non-overlapping homogeneous channels, each with same
bandwidth. A dedicated channel is utilized as common control channel for exchange
of control signals. The size of the beacon-period is smaller than the size of the active
period of primary users. A cognitive radio node can only perform out-of-band sensing
on other channels.
Each cognitive radio user maintains the spectral image of the primary user in a
vector defined by S I P. It also maintains a secondary channel load in a vector defined
by SC L. The S I P provides the local view of the spectrum and has following meaning
(i) (S I P[c] = 0) indicates there is no primary user active on channel c.
(ii) (S I P[c] = 1) indicates that there is a primary user active on channel c.
(iii) (S I P[c] = 2) indicates that the activity of primary user on channel c is uncer-
tain.
The S I P vector is used to determine if the network can user a certain channel for
data communication. The SC L vector is used for choosing the communication chan-
nel, it contains the expected load of the cognitive radio communication. The spectral
opportunity with the lowest SC L is picked when a node wishes to communicate.
During the AT I M window, the nodes participate in the frame synchronization
algorithm (for time synchronization), performing channel sensing, retrieving network
wide spectral opportunities, and performing two-way control handshake for data
exchange. The fast sensing is performed during the AT I M window while f ast
sensing is performed during the Data window. The f ast sensing is used to update
the S I P vector. If the S I P value of a channel is set to 2, that channel will be f ine
sensed. During the AT I M window, channel sensing results are also exchanged via
scan result packets (SRP).
In a multi-hop scenario, certain operations are additionally required for the correct
functionality of the cognitive radio system.
(i) A channel reservation is done during the AT I M window for utilization during
Data window via a 3-way handshake. In the AT I M frame, the sender inserts
its list of channel opportunities. Receiver compares this with his own list and
selects the channel with the least SC L values. The selected channels for data
communication is inserted into the AT I M − AC K frame. To mitigate hid-
den terminal problem, the transmitter rebroadcast the selected channel in the
AT I M − R E S frame.
(ii) To mitigate the problem of out-of-order frame delivery of AT I M, S R P and
beacon, a distributed coordination function is utilized. Different backoff win-
dows are applied for different type of traffic. Here traffic is divided into two
classes, (high priority and low priority). High-priority packets can select a back-
off timer w ∈ [0, C W/2), and low- priority packets have to select a backoff
timer w ∈ [C W/2, C W ). The value of backoff timer is reset each time a packet
is transmitted successfully. The beacon and S R P utilizes the high priority back-
off timer while the AT I M packet utilizes low priority backoff timer.
(iii) In a multi-hop scenario, the presence of a primary user should be detected by
all cognitive radio user even if they are operating on different channels. If a
secondary user sees the channel to be free during the current superframe and it
appeared to be busy during the last superframe then a deep channel sensing is
performed to check whether a channel is really free. This provides additional
insurance over primary users mitigation to interference. The time required for
vacating a channel is given as channel vacate time (CVT).
(i)
The miss detection probability of a secondary user is given by pmd . The false alarm
(i)
probability of a secondary user is given by p f a . The number of cognitive radio users
within the detectability range of the primary user is given as M. The probability of
randomly selecting a channel is given as 1/C. The network wide miss detection and
(n)
false alarm probabilities, pmd and p (n)
f a can be calculated as.
1 M
(n) (i)
pmd = 1 − (1 − pmd ) (2.41)
C
p (n) (i) N /C
f a = 1 − (1 − p f a ) (2.42)
A channel is only vacated if a cognitive radio detects at least one primary user on
the channel i.e. a primary user appears inside the detectability range of the cognitive
radio performing the channel sensing. The mean value of the CVT can calculated as:
∞
t B I (n) k
E[C V T ] = + ( pmd ) t B I [s]
2 k=1 (2.43)
tB I tB I
= (n)
− [s]
1 − pmd 2
where t B I is the length of the beacon interval.

A channel opening time is the time it takes to open up a channel on which a
primary user has ceased its operation. To simplify the analysis, it is assumed that the
false detection probability of cognitive radio is zero. If s (i) is the number of scans a
single cognitive radio requires to detect a channel as free, then it can be calculated as:
x−1
1 − p (n) C −1 p (n)
(i)
, x ∈ N+
fa fa
p(s = x) = + 0 (2.44)
C C C
Since, the system is utilizing a distributed sensing scheme, all cognitive radio need
to sense the channel as free before it is opened for communication. The number of
scans the entire cognitive radio network needs to detect the channel as free is denoted
by s (n) . Therefore,

x−1
p(s (i) ≤ x) = p(s (i) = k), x ∈ N+
0
k=1
(2.45)
M

M
p(s (n)
= x) = p(s (i)
= x) ∗ p(sm (i)
≤ x) M−m
, x ∈ N+
0
m=1
m
Hence the average channel occupancy time can be calculated as:

∞ t
BI
E[C O T ] = t B I x p(s = x) + [s] (2.46)
k=1
2
The primary user activity is modelled as Poisson process where the O N and O F F
periods are exponentially distributed with parameters μ O N and μ O F F respectively.
The probability that a channel is free is calculated as:
1
μ∗open
p f r ee = (2.47)
1
μ∗open
+ 1
μ∗closed
where μ∗open and μ∗closed are calculated as:

−1
1
μ∗open = + E[C O T ] − E[C V T ]
μO N
−1 (2.48)
1
μ∗closed = + E[C O T ] − E[C V T ]
μO F F
If all primary users are assumed to be inside the detection range of secondary users,
then the throughput of a single hop cognitive radio network can be calculated as:

C+1
St = C
Ci−1 p f r ee 1 − p nf a
i=1

C+1−i t B I − t AT I M (2.49)
∗ 1 − p f r ee 1 − p f a
n
tB I

∗ L − mod(L , i) Sc Li + mod(L , i)Sc Lii + 1
where L = N /2, denotes the number of communication links, x is the floor
function, mod(x, y) is the modulo operator, Sc (x) is the throughput estimation of
the given model with x as the number of links on the channel.
Increasing the number of cognitive radio nodes increases the false alarm proba-
bility and therefore there is a throughput loss. This can be compensate by increasing
the number of channels. Increasing the number of the cognitive radio nodes in the
vicinity of the primary users decreases the missed detection probability. This MAC
scheme utilizes distributed sensing scheme and solve the problem of hidden terminal
problem while providing the feature of broadcast and multicast. Since, this media
access scheme is based on dedicated control channel, it suffers from control channel
saturation problem. While contending over control channel the bandwidth is wasted
on data channels.
2.2.2 High Throughput Media Access Schemes
A cognitive MAC protocol (C-MAC) for distributed multi-channel is discussed in

[10], which utilizes distributed quiet periods for effective spectrum sensing. The
multi-channel is synchronized by utilizing varying beacon periods in different chan-
nels. It supports group communication like broadcast and multicast packet dissipation
through a common channel.
The protocol uses the concept of a rendezvous channel (RC) for coordination
between the nodes of the network. All the nodes agree on using exactly one channel
as RC when joining the network. Each channel time is divided into time-slots called
as superframes, which in turn has a beacon period and a data transfer period. The
beacon period across all channels are non-overlapping, which allows for efficient
utilization of the channel. Also, this means that if any node wants to get information
regarding other in all channels, the node has to switch channels taken in ascending
order of beacon period start time (BPST). On the other hand, for getting information
about a node and the channel it is located on, switching to the RC channel is adequate
as all the nodes periodically transmit their information beacons on the RC.
The RC is utilized to manage the entire cognitive radio network in the following
ways:
(i) Network wide group communication via broadcast and multicast packet trans-
mission.
(ii) Cognitive radio nodes visit the RC regularly to keep their clocks synchronized
so as they can adjust the channel beacon period start time. Also, this helps the
cognitive nodes to keep track of the offset they required to shift their BPST
when moving to other channels.
(iii) Each cognitive node visit the RC periodically and therefore can figure out the
other nodes residing on other channels. This helps to dissipate the neighbor
node information among all cognitive radio users.
(iv) During the exchange of beacons over RC, the number of beacons per channel
can be utilized to figure out the load on that channel. Here load refers to the
number of cognitive nodes residing on a channel.
(v) The RC also acts as entry point to any new cognitive radio node joining the
network. Any new cognitive radio node which wants to join the network first
searches the RC (with appropriate beacon header) and switches to it.
To join a network, a new node sends its beacon onto the RC letting other nodes know
about its existence. After learning the network wide information from the RC, the
new node switches to the appropriate channel and becomes part of that channel.
If no RC channel is found, the new node selects itself to make a decision on
marking a channel as RC. Once the node makes a decision on particular channel
as RC, it starts transmitting RC beacon on it, so that other cognitive users can join
it. It is possible that two or more cognitive user starts the process of starting RC
on some channels. In such case, convergence of RC is required. Cognitive nodes in
such system do out-of-band measurements of the spectrum. This is done to detect
the presence of other RCs and collecting channel quality information. If some other
overlapping RC is detected then the detector cognitive node moves back to its original
RC and transmit a RC-switch information in its beacon. This RC-switch information
informs other users on that RC to switch to the new RC along with the designated
time of moving. Other data channels are formed in a similar way by the cognitive
nodes, except that they send an additional beacon information over RC regarding the
channel switch information.
To dissipate the complete network information and mitigate hidden terminal prob-
lem, a form of distributed beaconing is employed in each channel by its occupants.
At the beacon period of each frame, each cognitive node in its allocated slot trans-
mits a beacon which contains information about its neighbours and neighbours of
neighbours (information received in the past frames). This allows all nodes to keep
track of other cognitive nodes on the same channel and allows mobility of cognitive
nodes on the same channel.
All the nodes periodically (and in a non overlapping fashion) switch to RC and
transmit their beacon for synchronisation purposes. This means that the information
contained in the RC is almost always up to date. This also helps in channel selection.
Any node in any of the channel can analyze the load on the channel from the number
of node beacons transmitted in the channel. This is rebroadcasted at the RC when
this node switches to RC during its resynchronization phase. Hence, other cognitive
nodes comes to know about the loads on other channels too. This allows selection
of appropriate channels by the cognitive nodes so as to decrease the contention and
increase the throughput.
Group communication like broadcast and multicast is tackled in the protocol using
the RC channel. Whenever a node has to perform a broadcast/multicast, it shifts to RC
and send a beacon scheduling a group communication. The group communication is
scheduled such that there is sufficient time in between for all other nodes to receive
the information from RC during their resynchronisation phase and be able to schedule
for group communication accordingly.
Another important aspect of this protocol is that it provide coexistence of sec-
ondary users with primary incumbents as well as other secondary users. The detection
of primary users is done during quiet period (QP) defined for all channels (and non-
overlapping for each). Once a primary user is detected, a beacon is transmitted to
notify other on the channel about the incumbent and channel evacuation is facilitated.
Coexistence with other secondary users (self-coexistence) is achieved via a beacon
period merging scheme. If an overlapping foreign BP is detected within the BPST
of the current cognitive node, the current cognitive node changes its BPST to the
foreign BPST and adjust its beacon slot accordingly. If a non-overlapping foreign BP
is detected, and its BPST lies within the first half of the current node’s superframe,
then the current node relocates its beacon to the foreign BP.
The scheme provides a robust coexistence mechanisms when incumbent nodes
are present in the channel. It provides proper mitigation of hidden node problem,
synchronization achieved in time, space and frequency via the beaconing approach.
The utilization of RC provide multicast and broadcast facility.
The scheme is dependent on RC for coordination making it a bottleneck. The
overhead for synchronisation and coordination becomes significant when data pack-
ets are small. Especially since a 3 way handshake is required in the data channel and
for a beacon period synchronisation in the beacon period of the frame. Due to the
requirement that the beacon period is non-overlapping, the frame length is a function
of the number of nodes on a channel and the beacon length. This may cause a signif-
icant degradation in throughput if the packets are small compared to the transmitting
window of the frame. The hidden terminal problem is addressed using neighbour
and next neighbour information included in the beacons transmitted. Mobility may
still cause nodes to go out of this extended awareness radius causing hidden node
problems.
2.2.3 Self-coexistence Based MAC Protocol
If cognitive radio based wireless devices make a breakthrough in practical day to

day life, then the opportunistic spectrum access will become more challenging. The
increase in complexity will be attributed to coexistence of multiple cognitive radio
networks in the vicinity of each other. Since one cognitive radio network will nor-
mally see other cognitive radio network as a foreign network or maybe as primary
incumbent of the channel on which it is residing. Most of the existing MAC proto-
cols in the literature takes a conservative approach in terms of spectrum sensing, i.e.
when a spectrum is sensed as busy, the cognitive radio node ceases to transmit. In
[11], a media access scheme is designed which considers the coexistence of multiple
cognitive radio networks and ensures fair access of resources among all users.
The problem with such an approach is that even if the channels are occupied by
other CR users from different networks, the node sensing the channel will still label
it as busy and will simply exclude it. Instead of utilizing a two state model (idle or
busy), a three-state sensing model for a channel is discussed.
In a system comprising of one primary user, M infrastructure based cognitive radio
networks (denoted as C R1 , C R2 , . . . C R M ) coexists. Each C Ri network comprises
of Ni secondary users. The cognitive radio node utilizes two transceivers: one for
communication and one is dedicated for spectrum sensing. Cooperative sensing is
done by nodes of the C Ri network and a decision is made regarding the channel
availability by the base station (BS) of the C Ri network. Let DT be the time required
for cooperative spectrum sensing, therefore the channel informations are updated
periodically every DT period of time. The channel sensed by the CR user (ri ) in a
the three state model is given as:
⎧
⎪
⎨n i H0
ri = x p + n i H1
⎪
⎩
xs + n i H2
where H0 , H1 and H2 are the state of channel being empty, channel occupied by a
primary user and channel occupied by some secondary user, respectively. x + p and
xs are the signals from the primary and secondary users, while n i is the zero mean
additive white Gaussian noise.
If a channel is busy, the cognitive node performs a fine sensing and analyze the
received energy in the second stage via distance estimation technique to identify
whether the channel is occupied by a primary user H1 or by some other cognitive
radio user H2 . When a cognitive radio user senses state H2 , it competes for the
channel in a fair manner among all remaining cognitive radio nodes as ensured by
the protocol. If instead it senses state H1 , it immediately vacates the channel to not
cause interference to the primary user transmission.
Here primary users have strict higher priority over secondary users. If a primary
user transmission is detected (H1 state), a secondary user immediately tries to vacate
the channel. This will require a certain duration of time, which appears as interference
time to the primary user. There is a bound on maximum interference time that can
appear to a primary user (Tmax ). If there is no primary user present on the channel
(H0 or H2 ), the secondary users continue accessing the channel to determine the
exact channel state (i.e. whether H0 or H2 ). If the channel state is determined as
H0 , the secondary user access the channels like any other cognitive radio node.
If the channel is determined in the H2 state, then the conventional IEEE 802.11
[12] distributed coordination function (DCF) mechanism is utilized for contention
and channel access as discussed in Sect. 1.5.1. If a secondary user has a packet for
transmission, it waits for a DCF interframe space (DIFS) duration before transmitting
the packet. If during this waiting period any other user is detected on the channel
then a backoff mechanism is utilized. The backoff duration is randomly selected
from [0, W − 1], where W is the size of contention window. To achieve the optimal
fairness among all secondary users, this method does not utilize binary exponential
backoff mechanism of IEEE 802.11.
If X represent a random variable denoting the time period during which a primary
user continuously transmits on a channel, its probability density function is repre-
sented as f X (x). Therefore the duration during which the channel is not accessible
(Y ) by any secondary user is given as:
X
Y = + 1 DT
DT (2.50)
≈ X + DT
The cumulative distribution of Y can be calculated as:

FY (y) = Pr (x + DT ) ≤ y
y−DT (2.51)
= f X (x)d x
0
Therefore, the mean of Y can be calculated as:

DT +t
E(Y ) = y FY (y)dy (2.52)
DT
If X follows an exponential distribution with parameter μ, the mean value of Y can

be calculated as:
DT +t
E(Y ) = yμe−μ(y−DT ) dy
DT
(2.53)
1
≈ + DT
μ
For a primary user following Poisson traffic with mean arrival rate as λ, the probability
that the channel is occupied during the time period t can be calculated as:
λt E(Y )
Poccupied =
t
1 (2.54)
= + DT λ
μ
Therefore the probability that the channel is free for secondary users to use is given
as P f r ee = 1 − Poccupied .
A Markov chain model is developed to do the analysis of throughput saturation
of the network. Where each state in the Markov chain represents the state of the
backoff stage of the secondary user’s transmission. Each time slot is considered
either a period of primary user occupancy, successful secondary user transmission,
idle or collision among multiple secondary user. The channel access probability Pα
is calculated as:
W
Pα = 1 − (2.55)
P f r ee (1 − Pα )N+ Poccupied + W
Therefore the probability that a transmission by secondary user over a given time
slot would be successful is given as:
N −1
Ps = P f r ee N Pα 1 − Pα (2.56)
The probability that a given time slot is idle slot is given as:
Pidle = P f r ee (1 − Pα ) N (2.57)
The probability that a collision happens in a given time slot is given as:

Pcs = P f r ee 1 − (1 − Pα ) N − N Pα .(1 − Pα ) N −1 (2.58)
If the average time of a successful secondary user transmission slot, the average time
of collision slot caused by other secondary users and primary users are represented
by Ts , Tcs and Tcp , then the throughput of secondary user is calculated as:
Ps Ts
θ= (2.59)
Pidle σ + Ps Ts + Pcs Tcs + Poccupied Tcp
where σ is the duration of an idle time slot.

The saturation throughput (θ∗ ) is calculated by optimizing the Eq. 2.59 as:
θ∗ = max{θ} (2.60)
This work primarily focused on coexistence technique including self-coexistence

which many other work fail to do. A kind of fairness scheme is utilized which
provides fairness to all cognitive radio nodes in the network. Extending this scheme
to an ad-hoc cognitive radio network would be tedious, as the system would have
to make distributed decision over channel state. These channel states may appear
different for different cognitive users within the same network.
A conceptual idea about an opportunistic spectrum access (OSA-MAC) [13]
scheme was discussed in Sect. 1.7.1. This protocol utilizes a dedicated control chan-
nel for control packet interchange ans coordination between the nodes. Multiple
opportunistic channels are utilized for data communication between secondary users
which are owned by primary incumbents.
The time is divided into superframes, which in turn are divided into ATIM window
and transmission intervals. During the ATIM window period, the secondary users
exchange ATIM (announcement traffic indication message) on the control channel
to indicate and agree on a channel for transmission. Once this has been done, the
participating nodes switch to the agreed channel during transmission interval, where
they compete with existing nodes using a RTS/CTS handshake and utilize the channel.
A preemptive sensing is done on the data channel before any further action to make
sure no primary users are present on it. If the case arises that primary users are
present, then the secondary users employ backoff periods and wait for the primary
users to vacate. The superframe period during second phase (where the secondary
senses the channel for primary user activity) is such that none of the secondary users
are allowed to transmit during this period (Quiet periods as discussed in Sect. 3.4.2).
The method utilizes two types of channel selection for data transmission: random
and opportunistic based. In a random channel selection a random channel is selected
with a probability of 1/L, where L is the total number of data channels utilized in
the system. In an opportunistic based channel selection, the spectrum availability
probability is taken into consideration. If the number of secondary users is small as
compared to the number of channels, it is better to allow the secondary users to exploit
more available channels instead of selecting any random channel. For a threshold
number of nodes at Nth and total number of users participating in communication
N , if N ≤ Nth the opportunity based channel selection approach is employed or else
the uniform channel selection method is employed. If pi, j is the probability that a
channel j is available for secondary communication i and if pi,k is the probability
that a channel k is available for secondary communication i, then the probability that
a user i choses channel j is given as:
pi, j
qi, j = L (2.61)
k=1 pi,k
The probability that k users selects channel j for data communication during trans-
mission phase is given as:
Ck, j = N Ck (q j )k (1 − q j ) N −k (2.62)
The probability that exactly one user wins the contention among k users on channel
j with maximum window size W is given as:
W
k l k−1
Sk, j = 1− (2.63)
W l=1 W
Therefore throughput on a channel j can be calculated as:
RtT I
N
Tj = Ck, j p j Sk, j (2.64)
t S F k=1
where R, tT I , t S F and p j are the transmission rate, time needed to transmit data on
each channel, length of the superframe and channel available probability of channel
j, respectively.
The total throughput for L channels is the summation of throughput on each
channel and can be calculated as:

L
T = Tj (2.65)
j=1
The probability of collision of a secondary user transmission with a primary user is

calculated via utilizing the probability of miss detection. The probability of collision
is thus given as:

N ! "
P j(c) = Ck, j 1 − (1 − Pm )k (1 − p j ) (2.66)
k=1
where Pm is the probability of miss detection by the given secondary user k, p j is

the probability of channel j to be free.
This opportunistic channel selection method provides a better saturation through-
put for cases where the total number of secondary is comparable to the number of
channels. There is an upper bound on the number of users a network can accommo-
date given a minimum QoS standard and the sensing error derived from it. The control
channel must be owned by the secondary service provider which would incur cost.
As the number of secondary users increase, the superframe period needs to increase
and since no transmission occurs on other channels during this period among the
secondary, this causes underutilization of resources.
2.3 Energy Management
Users in a cognitive radio network may be powered primarily by a portable battery

which generally has limited endurance between recharges. Therefore, energy man-
agement is an important issue that needs to be dealt in a cognitive radio network
for greener communication, extended operable life of a user between recharges and
to mitigate the effects of intermediate node failures [14]. Operating at low energy
forces a user to degrade its observed QoS, and hence a graceful degradation need to
be observed by such systems.
Energy management in a cognitive radio system depends upon multitude of things.
This can be performed at three different steps of cognitive radio network: transmis-
sion power control, spectrum decision and media access scheme. By doing transmis-
sion power control, a user limits the transmission power of a transmitter and hence
reduces the energy consumption. This reduces the SNR/SINR at the receiver side
and hence the observed QoS. Spectrum allocation based energy management tech-
niques employ selection of channels which require lower energy for transmission. A
media access scheme decides how a channel is shared among multiple users in the
network and thus its etiquette determines the activity of a user and hence its power
consumption. Depending upon cooperation between different secondary user in the
network, the sensing in the cognitive radio network can be classified into two types:
cooperative and non-cooperative.
2.3.1 Cooperative Sensing Based Energy Efficient Spectrum

Sensing
An energy efficient cooperative spectrum sensing scheme is discussed in [15], which

optimizes the sensing time in a given superframe satisfying the constraints of sensing
accuracy, throughput and latency. In a given superframe, there is a trade-off between
sensing and transmission. Similarly, there is also a trade-off between waiting on a
occupied channel to become free or switch onto another free channel.
If the sensing time is short then the sensing result is not accurate and hence detec-
tion of idle channel becomes impossible. Therefore carrying the transmission on
current channel without accurate sensing may result in throughput loss and interfer-
ence to primary incumbents. In other case, if the sensing time is too long then the
transmission duration becomes small. If the channel is sensed to be busy then the
secondary user can either switch to other channel or wait for the next time slot. If
the secondary user waits, it will conserve energy but it will also increase the trans-
mission delay and reduce throughput. Similarly, if it switches to another channel and
performs data transmission, the the throughput is increased at the cost of increased
energy consumption.
The time is divided into recurring superframes of equal length. Each superframe
length is divided into sensing and transmission duration. The system comprises of M
primary incumbent channels which are opportunistically utilized by the secondary
users. Each secondary user performs in-band sensing to detect whether the current
channel is idle or occupied. On the basis a decision is made on whether to stay on
the current channel or move to another idle channel.
2.3 Energy Management 73
As given in Eq. 2.91, the received signal at i-th secondary user is given as ri for
m-th sample, where s(m) is the primary users’ transmitted signal with zero mean
and σs2 variance while w(m) is noise which is modelled as a Gaussian process with
zero mean and σw2 variance (N (0, σw2 )); while H0 represents the hypothesis that the
channel is available while H1 represents the hypothesis that the channel is occupied
by the primary incumbents.
If γ is the average SNR of primary user, then the probability of detection (Pd ) and
false alarm (P f ) for sensing time τs is given as:

1 #
Pd (τs ) = Q √ Q −1 ( P¯ f ) − γ τs f s
2γ + 1
(2.67)
1 −1
#
P f (τs ) = Q √ Q ( P̄d ) + γ τs f s
2γ + 1
where Q(x) is the Gaussian tail probability function, while Q −1 (.) is its inverse.
f s is the sampling frequency. P¯ f and P̄d are target probability of false alarm and
detection, respectively.
The minimum sensing time to achieve the desired target probability of false alarm
and detection can be related as:
2
1 #
τsmin = Q ( P¯ f ) − Q −1 ( P̄d ) 2γ + 1
−1
(2.68)
γ 2 fs
Here the minimum sensing time is related to the desired target probability of false
alarm and detection. But this amount of sensing time may consume considerable
energy and an optimization over it is required constrained to maximum energy con-
sumption.
Based on the current channel sensing results, there exist three possibilities: current
channel is idle, all channels are busy, and current channel is busy with at least one
other channel is idle.
If the current channel is idle, then the secondary user makes a decision on to
stay on the current channel. Switching onto another channel is inefficient and will
result in increase energy consumption. If all M channels are busy (can be detected
by out-of-band sensing), then no channel switching is done. If the current channel is
busy and at least one another channel is idle, then a decision is made to minimize the
energy consumption. If the secondary user switch to another channel, then there is
increase in throughput at the cost of increased energy consumption. If the secondary
user stays on the channel, then the energy consumption is minimized at the cost of
throughput loss and increased latency.
If Ps is the probability that the secondary user stays on the current channel for
the above discussed third case, then a joint optimization problem can be formulated
with optimal sensing time (τs ) as following:
min E(τs , Ps )
τs ,Ps
subject to Pd (τs ) ≥ P̄d

P f (τs ) ≤ P¯ f (2.69)
W (τs , Ps ) ≥ W
D(τs , Ps ) ≥ D
where E is the total average energy required to transmit one packet, W is the average
throughput, D is the average delay, W is the minimum average throughput and D is
the maximum average delay.
If E s and E t represents energy of sensing and transmission in each time-slot
respectively, then the total average energy consumption can be given as:
E(τs , Ps ) = N τs E s + N Psw E sw + Ttr E t (2.70)
where N is the total number of slots needed for one data packet transmission, E sw
is the energy cost for one channel switching, Psw is the probability of switching to
idle channel, Ttr is time required for one packet data transmission.
The average throughput W can be calculated as a function of τs and Ps with the
time-slots of period T as:
Bst + Bsw
W (τs , Ps ) = (2.71)
τs + T
where Bstay is the average number of bits that are transmitted in the duration τs + T
when the secondary user chooses to stay on the same channel; while Bswitch is the
average number of bits that are transmitted in the duration τs + T when he secondary
user chooses to switch to another channel.
Bst and Bsw can be given as:
Bstay = Pidle BT
(2.72)
Bswitch = Pswidle BT (1 − Pswbusy )
where Pidle is the probability that a channel is correctly sensed by the secondary
user, Pswidle is the probability that a secondary user switches to an idle channel while
Pswbusy is the probability that a secondary user switches to a busy channel.
The average delay can also be expressed as a function of τs and Ps with the
time-slots of period T as:

D(τs , Ps ) = N τs + T Pwait (2.73)
where Pwait is the probability that the secondary user makes a decision of staying on
the current channel and N is the number of frames required for completing the data
transmission.
From Eq. 2.69, an optimal value of Ps can be achieved for energy minimization
via setting, W (τs , Ps ) = W and D(τs , Ps ) = D. Therefore the optimization problem
in Eq. 2.69 is reduced from joint optimization to a single optimization variable τs .
The system assumes that all the channels are sensed at the same time with same
accuracy which is not true and thus the probability of false detections and finding busy
channels are affected. Especially in the heterogeneous environment where channel
characteristics varies.
In conventional distributed sensing approach, the higher detection performance
comes with a higher network energy consumption. In [16], a distributed sensing
technique is discussed which reduces the network energy consumption by optimally
choosing the sleeping and censoring parameters. The energy consumption is con-
strained and subjected to the the minimum detection performance and a maximum
false alarm probability.
The system comprises of N cognitive sensor nodes and a fusion centre (FC) in a
parallel distributed fusion configuration. In a parallel distributed fusion configuration,
each cognitive sensor node makes a decision on the channel availability and sends
the result to fusion centre. The fusion centre then fuses these local sensing data and
makes a decision on the channel availability. Each cognitive radio node follows two
policies: sleeping policy and censoring policy. The sleeping policy decides whether
a cognitive radio node is awake or in sleep mode, while the censoring policy decides
whether a sensed results is transmitted or not.
As given in Eq. 2.91, the received signal at the m-th sample for the i-th secondary
user is given as ri . Also, SNR at each cognitive radio node is considered to be
same (γ). The cognitive radio node performs binary hypothesis on energy detection.
H0 represents the hypothesis that the channel is available while H1 represents the
hypothesis that the channel is occupied by the primary incumbents. s(m) is the
primary users’ transmitted signal with zero mean and σs2 variance while w(m) is a
Gaussian process with zero mean and σw2 variance (N (0, σw2 )). Therefore the received
energy at the i-th cognitive radio node over T0 observation samples is given as:

T0
Ei = ri2 (k) (2.74)
k=1
A form of thresholds (λ1 , λ2 ) are applied to cognitive sensor nodes to enable cen-
soring policy. The threshold bound λ1 < E i < λ2 , is called as the censoring region.
Therefore by this rule, the local decisions made by the cognitive sensor nodes can
be given as:
⎧
⎪
⎨send 1 (implying H1 ), if E i ≥ λ2
result = no decision, if λ1 < E i < λ2 (2.75)
⎪
⎩
send 0 (implying H0 ), if E i ≤ λ1
The false alarm probability (P f ) and detection probabilities (Pd ) can be given as:
(T0 , λ2 /2)
Pf =
(T0 ) (2.76)
# #
Pd = Q T0 ( 2γ, λ2 )
$where (n, x) is an upper incomplete gamma function given by (n, x) =

∞ n−1 −t
x t e dt, with (n, 0) = (n), and Q u (n, x) is the generalized Marcum Q-
$ ∞ − t 2 +n2
function given by, Q u (n, x) = n u−1
1
x e
2 Iu−1 (nt)dt. Iu−1 (.) is the Bessel func-
tion of the first kind with the order u − 1.
The cost function is modelled as the average energy consumed by each cognitive
radio node in the system, which can be given as:

N
C T = (1 − μ) (Csi + Cti (1 − ρ)) (2.77)
i=1
where, Csi and Cti are the energy required for sensing and transmission by i-th
cognitive sensor node, respectively. μ is the sleeping rate while ρ is the censoring
rate and equals to the probability such that λ1 < E i < λ2 .
The constraints are applied over the global detection (Q D ) and false alarm (Q F )
probabilities. Therefore the optimization problem can be formulated as:
min C T
μ,λ1 ,λ2
such that Q D ≥ β (2.78)

QF ≤ α
where β and α are parameters that can be modified for different use case scenario.
Depending on the prior knowledge about the probabilities of hypotheses H0 and
H1 , two different cases are discussed.
2.3.1.1 Unknown Prior
In case of unknown prior, it can be assumed that Pr (H0 ) ≪ Pr (H1 ) (probability of

hypothesis H0 is very less than hypothesis H1 ) implying that channel is mostly free
from primary incumbents. The problem then can be formulated as blind Neyman–
Pearson setup, which can be given as
ρ N eyman = Pr (λ1 < E i < λ2 |H0 ) (2.79)
From Eq. 2.76, the above equation can be rewritten as:
(T0 , λ21 ) (T0 , λ22 )

ρ N eyman = − (2.80)
(T0 ) (T0 )
From Eqs. 2.77 and 2.80, the Eq. 2.78 can be given as:
N
(T0 , λ21 ) (T0 , λ22 )
min (1 − μ) Csi + Cti 1 − +
μ,λ1 ,λ2
i=1
(T0 ) (T0 )
(2.81)
such that Q D ≥ β
QF ≤ α
The global probability of false alarm (Q F ) in the Neyman–Pearson setup can be

calculated as:
N K
N N −k K K −l
QF = μ (1 − μ)k ∗ ρ (1 − ρ)l [1 − (1 − P f )l ] (2.82)
k=1
k l=1
l
Using binomial expansion theorem this can be reduced to

! "N
Q F = 1 − 1 − P f (1 − μ)(1 − ρ) (2.83)
The global probability of detection (Q D ) in the Neyman–Pearson setup can be cal-

culated as:
N K
N N −k K K −l
QD = μ (1 − μ) ∗
k
δ (1 − δ)l [1 − (1 − Pd )l ] (2.84)
k=1
k l=1
l
Using binomial expansion theorem this can be reduced to

! "N
Q D = 1 − 1 − Pd (1 − μ)(1 − δ) (2.85)
Therefore by utilizing the value of Q F and Q D , the optimization problem in Eq. 2.81
can be solved to derive optimum sleeping rate (μ) and censoring thresholds (λ1 , λ2 ).
2.3.1.2 Known Prior
If the probabilities of hypothesis H0 (Pr (H0 )) and hypothesis H1 (Pr (H1 )) are
known in advance, then the problem can be formulated as knowledge aided Bayesian
setup. Therefore the censoring rate can be given as:
ρ Bayesian = π0 Pr (λ1 < E i < λ2 |H0 ) + π1 Pr (λ1 < E i < λ2 |H1 ) (2.86)
where π0 and π1 are the probability of hypothesis H0 and H1 , receptively.

From Eq. 2.76, we have:
(T0 , λ21 ) (T0 , λ22 )

Pr (λ1 < E i < λ2 |H0 ) = −
(T0 ) (T0 ) (2.87)
# # # #
Pr (λ1 < E i < λ2 |H1 ) = Q T0 ( 2γ, λ1 ) − Q T0 ( 2γ, λ2 )
From Eqs. 2.77 and 2.87, Eq. 2.78 can be rewritten as:
min (1 − μ)
μ,λ1 ,λ2
N (T0 , λ1 ) (T0 , λ2 )
Csi + Cti 1 − π0 2
− 2
i=1
(T 0 ) (T 0 ) (2.88)
# # # #
+ π1 Q T0 ( 2γ, λ1 ) − Q T0 ( 2γ, λ2 ) such that Q D ≥ β
QF ≤ α
The local probability of false alarm and detection in knowledge aided Bayesian setup
can be given as:
P fB = (1 − μ)(1 − δ0 )P f
(2.89)
PdB = (1 − μ)(1 − δ1 )Pd
Therefore, the global probability of false alarm and detection in knowledge aided
Bayesian setup can be derived as:
! "N
Q F = 1 − 1 − P f (1 − μ)(1 − δ0 )
! "N (2.90)
Q D = 1 − 1 − Pd (1 − μ)(1 − δ1 )
where δ0 = Pr (λ1 < E i < λ2 |H0 ) and δ1 = Pr (λ1 < E i < λ2 |H1 ).
Therefore by utilizing the value of Q F and Q D , the optimization problem in
Eq. 2.88 can be solved to derive optimum sleeping rate (μ) and censoring thresholds
(λ1 , λ2 ).
The blind Neyman–Pearson setup is shown to be an special case of Bayesian
setup. As the transmission energy increases compared to the sensing energy, the
total transmission energy has to be reduced more than the sensing energy thus the
censoring rate increases and the sleeping rate decreases. It is shown that without
sleeping or censoring, the energy consumed saturates to a level which is much lower
to conventional energy consumption. Also, as the number of users increase, the
optimal sleeping rate increases dramatically to keep the energy consumption of the
system stable.
2.3.2 Non-cooperative Sensing Based Energy Efficient

Spectrum Sensing
Traditionally usage of cooperative sensing provides efficient channel occupancy

results in a resource constrained environment, but it requires exchange of control
messages. This exchange of control messages for cooperation increases the energy
consumption of cognitive radio network. In [17], a time division energy efficient
(TDEE) non-cooperative spectrum sensing scheme for cognitive radio network is
discussed. In this scheme, the sensing period is divided in time slots and each sec-
ondary user in the network is assigned a different channel to detect in each slot. This
allows secondary users to accurately sense the network without exchanging control
messages for cooperation and allowing to save considerable energy.
The system utilizes a central controller node (base station in infrastructure net-
work or cluster head in case of ad-hoc network) which coordinates the cognitive
nodes in the network. The central controller designates cognitive nodes in the net-
work to perform sensing on given channels along with the given sensing order. The
information is then received by the central controller from the cognitive nodes, which
in turn perform fusion on the received spectrum sensing data to make a decision on
the availability of channels. The system time is divided into superframes, where
each superframe comprises of two parts: sensing and transmission. The duration of
the superframe is given as T , in which Ts is the time duration of spectrum sensing
whereas Tr is the time duration of transmission in the superframe. The time Ts is
further divided in equal time slots during which the secondary user senses different
channels in the order given by the central controller. After the expiration of sensing
period all the SUs send the sensing results to the central controller. For given N
secondary users and M channels in the system, the given scenario considers where
U secondary users want to perform sensing, where 2 ≤ U ≤ min(N , M).
The steps required for a secondary user to perform the sensing in the discussed
scheme can be given as:
(i) All secondary users join to the network controller by central controller by fol-
lowing their network join/leave protocol. Once joined, it waits for the control
messages from the central controller.
(ii) The central controller provides the information on number of channels to be
sensed by the secondary users. Depending upon that, the sensing duration is
divided into mini time-slots during which the secondary user individually scans
each channel. Once the sensing is performed on all channels, the secondary
users will update the information to the central controller.
(iii) The central controller fuses the channel availability information received from
all secondary users and based on that selects a channel.
Since, the control packets are only exchanged with central controller, the number
of control packets exchanges is substantially reduced. The efficiency of the discussed
can be analysed in two schemes: Homogeneous networks and Heterogeneous net-
works.
2.3.2.1 Homogeneous Networks
In a cognitive radio based homogeneous network environment, all secondary users

have the same sensing time and same channel coefficient h from primary incumbent’s
transmitter. The sensing duration is divided in U time slots and length of each slot
is τ such that Ts = U τ .
Total energy spent in the sensing duration is sum of energy spent for sensing
different channels and energy used for exchange messages. It is clear that the sens-
ing/exchange energy will not only depend on the sensing/exchange power but also on
sensing/exchange time. Although the exchange duration is very small the exchange
energy can be neglected, but as the number of cooperative secondary users increase,
the exchange energy will also increase (keeping in mind retransmission if error hap-
pens). Therefore the energy consumed via exchange of control packets is also taken
into consideration. The received signal ri at the m-th sample for the i-th secondary
user can be given as:

w(m), H0
ri (m) = (2.91)
s(m) + w(m), H1
where H0 represents the hypothesis that the channel is available while H1 represents
the hypothesis that the channel is occupied by the primary incumbents. s(m) is the
primary users’ transmitted signal with zero mean and σs2 variance while w(m) is a
Gaussian process with zero mean and σw2 variance (N (0, σw2 )).
The detection probability can be calculated by utilizing the central limit theorem
over the Chi-square distribution and is given as:

λ − τ f s (|h|2 σs2 + σw2 )
Pi = Q √
d
(2.92)
2τ f s (|h|2 σs2 + σw2 )
Similarly, the false alarm probability can be calculated as:

f λ − τ f s σw2
Pi = Q √ (2.93)
2τ f s σw2
$ ∞ t2
where, Q(x) = √12π x e− 2 dt is the tail probability of the standard Normal distri-
bution function and f s is the sampling frequency.
The detection probability and false alarm probability calculated by the central
controller can be given as:
P d (U ) = 1 − (1 − P d )U
(2.94)
P f (U ) = 1 − (1 − P f )U
If Pt x and Pr x are the transmitted power and received power at each secondary user,
respectively, then the optimization problem of calculation of requirement of number
of secondary user for energy minimization can be calculated as:
! "
min :U Ts Pt x + Te Pr x
U
such that, P d (U ) ≥ PTdh (2.95)

f
P f (U ) ≤ PT h
where Ts Pt x is the sensing energy consumed in the homogeneous network and Te Pr x

is energy consumed by exchanging control messages in which Te is the control
message exchange time. Pd,T h is the threshold of the cooperative detection and P f,T h
is threshold of false alarm.
By applying bounds on the constraints of Eq. 2.95, the bounds on U can be given
as:
f
ln(1 − PTdh ) ln(1 − PT h )
≤U ≤ (2.96)
ln(1 − P )
d ln(1 − P f )
Since, the optimization Eq. 2.95 is a decreasing function of the parameter U , the
minimal value of U for minimal energy consumption is given as:
ln(1 − PTdh )
U= (2.97)
ln(1 − P d )
If Ps is the probability that a channel is successfully found then it can be given as:
! "
Ps = p 1 − P f (U ) (2.98)
where p is the channel available probability in a 2-state Markov chain model of

spectrum occupancy.
Therefore the probability that the available channels can be found in one sensing
period (Pavailable ) can be given as:
U
U
Pavailable = (1 − Ps )U −u Psu (2.99)
u=1
u
The throughput of a channel can be derived by utilizing the Eq. 2.99:
TH = (T − U τ )Pavailable R (2.100)
where R is the bandwidth of the channel.

2.3.2.2 Heterogeneous Networks
In heterogeneous networks all secondary users have different sensing periods and
different channel coefficients (h i , ∀i ∈ U ). In this case the total energy consumed
will be dependent on the number of secondary users (U ) along with the sensing time
of each secondary user. If τi is the size of the time slot of i-th secondary user, then
the optimization equation for minimizing the energy consumption can be given as
U !
"
min Tsi Pr x + Te Pt x
U ;Ts1 ,...TsU
i=1
(2.101)
such that, P d (U ) ≥ PTdh
f
P f (U ) ≤ PT h
U
The Eq. 2.101 can be refined for the condition that i=1 Tsi is minimized when all
Tsi are minimized and therefore can be rewritten as:
min Tsi = U τi
U ;τi
such that, P d (U ) ≥ PTdh (2.102)

f
P (U ) ≤
f
PT h
A bound on the value of U can be calculated as:

f
ln(1 − PTdh ) ln(1 − PT h )
≤U ≤ (2.103)
ln(1 − Pi,min )
d f
ln(1 − Pi,max )
The optimization problem of finding the optimal sensing timeslots can be given as:

λ−τ f (|h |2 σ 2 +σ 2 )
Q √2τ f s (|h min |2 σs2 +σw2 )
ar g max w
s min s
τi τi
f
(2.104)
ln(1 − PT h )
such that, U ≤ f
ln(1 − Pi,max )
The optimal value of the sensing time slot is calculated as:

%
∗ 1 λ2
τi = −1 (2.105)
2 fs 4(|h min σs2 + σw2 ) + 1
Similar to the homogeneous environment case, the throughput in heterogeneous

environment can be calculated as:

U
TH = T − τi Pavailable R (2.106)
u=1
where, R is the bandwidth of each channel and Pavailable is the probability that the
available channels can be found in one sensing period.
The energy consumed by this approach is always less than the traditional coop-
erative scheme. The throughput of this scheme is much higher than the traditional
scheme as during the same sensing duration, the number of channels discovered
correctly is always more.
A distributed resource allocation based energy efficient scheme is discussed in
[18], for cognitive radio based wireless sensor network operating in multi-carrier
environment. The discussed scheme allows subcarrier selection and power allocation
among individual cognitive radio nodes. The overall goal of the scheme it to reduce
the energy consumption per bit over all subcarriers while maintaining the certain
level of data transmission rate.
Multiple users are allowed to share the same subcarrier as long as their Signal
to Interference and Noise Ratio (SINR) is acceptable. The problem of subcarrier
(channel) allocation is converted into an unconstrained optimization problem and
then branch and bound method of solving optimization equations is utilized to find
an optimal solution.
As discussed in Sect. 1.7.2, the system time is divided into time slots where the size
of each time slot is Ts . During each time-slot the cognitive nodes exchange beacons
which is utilized for synchronization purposes. The entire frequency spectrum is
divided into M subcarriers of which each spectrum experience flat Rayleigh fading.
After each user senses the network and obtains the list of available subcarriers,
the scheme will select subcarriers and allocate power to them such that the energy
efficiency is maximized, and data rate and power constrains are satisfied. If Pt(i) is
the transmission power allocated to sub carrier i, then the optimization problem of
power allocation in distributed subcarrier can be given as:
M
Pt(i) + Pr
i=1
min M (i)
(2.107)
Pt(i) B. i=1 log2 1 + αi Pt
Subject to,

M
M

R (i) = B. log2 1 + αi Pt(i) ≥ Rr eq , ∀i = 1, 2, . . . , M
i=1 i=1
(2.108)

M
Pt(i) = Pmax , Pt(i) ≥ 0, ∀i = 1, 2, . . . , M
i=1
where Pr is the total circuit power consumption, B is bandwidth of each subcarrier,

Pmax is the maximal power bound, Rr eq is the target data rate, αi is channel state
information of subcarrier i and M is the number of available channels.
The optimization Eq. 2.107 is not quasi-concave/convex and therefore to solve it,
the constraints from this Equations are removed initially. Once the optimal solution to
unconstrained problem is known, the initial constraints of Eq. 2.108 can be examined
again.
On removing
∗
the constrains from the Eq. 2.107, the optimal transmission power
P ∗ = Pt(i) , ∀i = 1, 2, . . . M can be given as [18]:
1
(i)∗ e ∗
Pt = max B.log2 .ζ − ,0
αi
M (i) (2.109)
i=1 Pt + Pr
ζ ∗ = min∗ M (i)∗
Pt(i) B. i=1 log2 1 + αi Pt
where ζ ∗ is optimal energy per bit.

Multiple solutions exist for Eq. 2.109. The constraints of Eq. 2.108 are examined.
For two constraints, a total of four possibilities exists:
(i)∗ (i)
(i) If P ≤ Pmax and R ≥ R , then Eq. 2.109 is satisfied.
t (i)∗ (i) r eq
(ii) If Pt ≥ Pmax and R ≤ Rr eq , it implies that even after allocating
all the channels which exceeds the maximal power bound, the target data rate
requirement is not met. Therefore in such condition, there exist no feasible
solution
(i)to∗ the problem. (i)
(iii) If Pt ≤ Pmax and R ≤ Rr eq , it implies that the target data rate
requirement is not met. Since power allocated to subcarrier still hasn’t reached
the maximal bound, power should be increased till the maximal power bound
is reached. The minimal required additional power to achieve the data rate of
Rr eq can be given as:
& M (i)∗
'
Rr eq − i=1 R
Pmin = Mβ e Blog2e M
−1 (2.110)
(i)∗
If even after this additional power, if Pt + Pmin ≥ Pmax , then there exist
no feasible∗
solution to the Eq.
(i) 2.107 with constraints in Eq. 2.108.
(iv) If Pt(i) ≥ Pmax and R ≥ Rr eq , it implies that total power allocation
and data rate both exceeds their upper bound. In this case, the power allocated
in subcarrier is decreased so as to make it equal to or below the Pmax value.
The optimal subcarrier selection and power allocation is done independently by all
new users, which may lead to multiple users deciding to use the same sub-carrier
and thus causing interference on it. Therefore a power control scheme based on
distributed co-channel interference mitigation is required. If N (i) new users want
to use the same subcarrier i at the same time, then the following distributed power
control iteration will converge if a feasible solution exist.
r eq,(i)
γn
Pn(i) (k + 1) = Pn(i) (k) (2.111)
γn(i) (k)
Here Pn (i)(k), γn(i) (k) are transmission power and measured SINR of new user n on
subcarrier i in step k, respectively. γ r eq,(i) is the target SINR and is calculated from
the optimal allocation data rate Rn(i)∗ .
Each node only needs to know its received SINR at the receiver and use it to change
its transmission power. The receiving node returns this as the feedback through a
control channel. If the power control algorithm cannot resolve the interference issues
then it should be resolved by the medium access control (MAC) scheme which needs
to be developed. Each user can only detect interference from other user after they
have started transmitting. Also it may take a lot of steps for the iterative algorithm
to converge which can waste a lot of time.
In [19], an energy efficient source and channel sensing (JSCS) scheme is discussed
for cognitive radio sensor networks. The overall work of the scheme is divided into
two tasks: ambient sensing for vacant channels and application sensing for collecting
application oriented information and delivering it to the access point/gateway. Both of
these tasks consume ample amount of energy, therefore the objective of the discussed
scheme is to reduce the total energy consumption while maintaining a bound on the
distortion of the application-specific source information.
The ambient oriented channel sensing performs periodic channel sensing so as
to detect the available channels for transmission. This helps to detect and deter-
mine the energy distribution of the spectrum and share the information with other
cognitive devices in the network for cooperation. The application oriented source
sensing involves invoking a sensor in the cognitive sensor node (sensors like temper-
ature, humidity, luminosity etc.) and passing the obtained information to the access
point/gateway.
The cognitive sensor network perform the two tasks of channel sensing and appli-
cation sensing simultaneously. The objective of the system is to minimize total con-
sumed energy in the system by doing a trade-off between the energy consumed
while doing the channel sensing and while doing the application sensing. If exces-
sive resources are given to application sensing then even after the collection of precise
application information, it cannot be transmitted in due time due to the lack of chan-
nel sensing information. If channel sensing is given excessive resources then channel
availability can be determined reliably but then the application related information
might not be delivered at an acceptable rate to the access point. Thus there exist a
trade-off between these two tasks.
The system time is divided into recurring superframes. The time period of each
superframe is T and is short enough for a primary incumbent activity to remain
unchanged in this timeframe. At the start of each superframe each node in the cogni-
tive sensor node performs channel sensing via detection of N samples on the spec-
trum. The cognitive sensor network operates on carrier frequency f c and bandwidth
W and performs N sample of this channel at the starting of each superframe.
The received signal at any cognitive node can be expressed as shown in Eq. 2.91.
Here r (m) is the received energy at m-th sample, with s(m) as the primary users’
transmitted signal with zero mean and σs2 variance while w(m) as a Gaussian process
with zero mean and σw2 variance. Therefore the energy detector for N samples can
be given as:
1
N
T (r ) = |r (m)|2 (2.112)
N m=1
For sufficient number of samples, the probability density function (T(r)) via central
limit theorem can be expressed as:

N (μ0 , σ02 ), H0
T (r ) = (2.113)
N (μ1 , σ12 ), H1
As discussed in Eq. 2.91, H0 and H1 are hypothesis for the absence and presence of
the primary user on the given channel, respectively. Here the primary user transmitted
signal is modelled as MPSK complex signal.
If p f alse and pdetect are the probability of false alarm and detection respectively,
then the probability of missed detection is given as, pmiss = 1 − pdetect , and the
probability that the cognitive node is allowed to transmit on a given superframe is
calculated as:

pt = 1 − p f alse p(H0 ) + 1 − pdetect p(H1 ) (2.114)
where p(H0 ) and p(H1 ) are inactive and active probabilities of primary incumbents.
If collision probability is given as pcollision = pmiss p(H1 ), then the effective
transmission probability for a cognitive node is given as:
pt = pt − pmiss p(H1 ) (2.115)
If the energy consumed by the detection on each sample is denoted as E sample , then
the average channel sensing power consumption on a single superframe can be given
as:
E sample N
Pch−sensing = (2.116)
T
The transmission power with respect to effective transmission probability pt can be
given as:
⎡ ( ⎤2
2σs2
Q −1 1 − pt
− + 1Q −1 (1 − pmiss )
⎢ p(H0 ) σw2 ⎥ E sample
Pch−sensing ( pt ) = ⎣ σs2 ⎦
T
σw2
(2.117)
−1
where Q is the inverse tail probability of the standard normal distribution.
The application sensing power comprises of energy consumption in sensor acti-
vation, source-channel coding and transmission. The energy required for correct
delivery of each bit can be given as:

2(Rchannel /W ) − 1
E bit = N0 W (2.118)
Rchannel
where Rchannel , N0 and W are maximum channel capacity, unilateral noise power
spectral density and channel bandwidth, respectively.
Therefore the average application sensing power can be given as:
Papp−sensing ( pt ) = pt E bit Rchannel (2.119)
For a multi-terminal source coding system, the Eq. 2.119 can be rewritten as:
σ2
( Ds )1/K
2 pLt W
Papp−sensing ( pt ) = N0 W pt + pmiss p(H1 ) σ2 −1
1 − Kw D1 − 1
σs2
(2.120)
where K is number of nodes and D is distortion. σs2 and σw2 are the variance of the
source and the variance of noise, respectively.
If the total power of a cognitive node is constrained then a trade-off is done between
the application sensing and channel sensing. The effective transmission probability
is the key that connects the two parameters. Therefore a joint transmission model is
considered where total energy consumed can be given as:
Ptotal ( pt ) = Papp−sensing ( pt ) + Pch−sensing ( pt ) (2.121)
If the probability of false alarm is kept below 0.5 then Ptotal ( pt ) forms a convex
curve with respect to pt and therefore an optimal solution exist for minimal total
power consumption. By utilizing the Eq. 2.121, an optimization can be done over
the application sensing and channel sensing power so as to satisfy a given distortion
constraint of the source coding.
Table 2.4 Comparison of simulation, emulation and real-time test beds

Simulation Emulation Real-time test beds
Implementation Based on probabilistic Based on simulated Based on realtime
and analytical and some real system components
modelling of components
components
Behavior Results are Results are quite Results are realistic
macroscopic and accurate and
moderate in terms of near-realistic
accuracy
Replication Replication is Replication to near Replication can be
improbable due to value can be achieved achieved via utilizing
varying parameters the same components
Realization Minimal Modest High
complexity
Computation Very high High Minimal
complexity
Cost Minimal Modest Expensive
2.4 Cognitive Radio Platforms
The performance of a cognitive radio system may be evaluated via either three
method of implementations: simulation, emulation and real-time test beds. Simu-
lation allows performance analysis of the overall system into an abstract level while
providing macroscopic facts and figures. Emulation involves utilization of real-time
(real-system) component integrated into a simulation system thereby providing a
more accurate form of analysis. The small-scaled deployment of a real-time test-bed
allows providing realistic results and accurate microscopic behavior of the overall
system [20]. Each of these techniques have their pros and cons. Table 2.4 provides
pros and cons of each of these techniques.
In wireless environment, simulation model normally misses the complex channel
characterization. The wireless channels are complex in terms of shadowing, fading,
noise,1 interference and multi-path propagation. Collective modelling of these para-
meters poses deviation from realistic behavior which can only be solved via using
realistic components in the evaluation system. Emulation tend to solve this problem
by utilizing realistic components instead of complex models in the evaluation system.
This provides more realistic results at the cost of utilization of realistic components.
To provide accurate realistic results, real-world platform need to be deployed as
test-beds [21, 22]. Therefore development of platforms for test-bed is important for
evaluation of cognitive protocols.
1 Here we refer noise as intrinsic noise, which is generated by the communication device itself;
while we refer interference as the extrinsic noise, which the communication device receive from
other unintended signal sources.
2.4 Cognitive Radio Platforms 89
There has been various advances in realizing a real-time, at-scale validations of the
cognitive radio technologies [22–24]. To realize such platform, generally a frequency
agile flexible radio is preferred which is capable of generating multitude of waveforms
controlled via software. This is normally achieved via a software defined radio. These
platforms tend to accommodate design and development specifications that cover up
the current scenario and scenarios yet to come. Most of the platforms are designed to
be modular, so as to allow ease in development, debugging and multiple iterations of
verification. The modular design also allows the platform to be structured and clean
slate for easy learning and additional component building.
2.4.1 From FPGAs to Software Defined Radio
One of the requirements is that cognitive radio systems change its behaviour accord-
ing to its context autonomously, based on experience and learning. Therefore the
system should be able to interact with the reconfigurable network node. Also, since
real time sensing is computationally intensive, it requires for high demanding digital
signal processing algorithms.
Until recently, reconfigurable logic and Field Programmable Logic Array (FPGA)
chips have been used primarily for prototyping. However, it is now being realized that
reconfigurable architectures are a separate family in their own right, just as proces-
sors and Application specific Integrated Circuits (ASIC) are with unique properties
that can be taken advantage of. General Purpose Processors are highly program-
mable platforms, however the performance is limited. The basic building block of
FPGA is a reconfigurable cell which is implemented using Look-Up-Table (LUT),
Programmable Logic Array (PLA) or Memory. LUT or memory-based cell is con-
figured by writing the result of the combinational logic to the cell [25]. That is some
modules need to be selectively reconfigured or reassigned for new mode of operation
in software defined radio. Therefore, reconfigurable architectures will give greater
flexibility in such an environment [24].
2.4.2 From Software Defined Radio to Cognitive Radio
Cognitive radio was first conceptualized from software defined radio by J. Mitola
while his work on the field of cognitive radio [26]. The capabilities of SDR to
tune to any frequency, selecting any band filter, generation of any modulation via
software allows the realization of cognitive radio. It is well known that CR are
computationally intensive systems that implement highly demanding digital signal
processing algorithms on different platforms. Design of mutli-mode platforms using
FPGA based SDR would be ideal for cognitive radio architectures. The possibility
of such a system has been investigated in detail [27]. A comparative study of SDR
based cognitive radio platform is done in [23].
GPU/CPU
FPGA
Application Control
Layer processor
Network RF
Layer Front End
FIFO
FIFO
MAC Base Band processing
Fig. 2.3 Block schematics of SDR
The component therefore which translate SDR into Cognitive Radio is Cognition.
Apart from cognition, other software components like protocol stacks are required
to realize the full functioning cognitive radio network. In the next section we will
discuss about different SDR control software that are utilized to emulate the cognition
behavior.
Figure 2.3 shows a basic block schematic of Software defined Radio (SDR) archi-
tecture. Fundamentally, the idea of SDR is to replace most of the analog signal
processing in the transceivers with digital signal processing in order to provide the
advantage of flexibility through reconfiguration or reprogramming. This will enable
different air-interfaces to be implemented on a single generic hardware platform. A
multi-mode SDR handset with dynamic reconfigurability has the promise of inte-
grated services and global roaming capabilities. In order to become reality, the SDR
technology requires more research not only from the perspective of wireless systems,
but also in the areas of novel computing architectures, embedded systems, and design
methodologies to realize an all-digital reprogrammable radio.
2.4.3 SDR Software
Software controls are required for providing control over the SDR hardware. These
software controls can be utilized to program a SDR to behave like a radio with
cognition. Some of these SDR software are: GNU Radio Software, Pothosware,
IRIS, OSSIE [23].
Figure 2.4 shows the block diagram of an SDR with an application software con-
trolling it. The application software controls the radio behavior including frequency
selection, modulation, bandwidth, gain, noise filtering etc.
Fig. 2.4 Application

software on top of SDR Software
Controller interface
Signal A/D, Multiband

Antenna
Processing D/A RF
2.4.3.1 GNU Radio
GNU Radio is a free, open source development toolkit for SDR [28]. It is licensed
under GNU GPL version 3. It is widely used in academia and in commercial ecosys-
tem. The toolkit comprises of different modules each representing an individual step
in signal processing. The overall functionality of the toolkit is achieved by appli-
cations which renders themselves as a ‘flowgraphs’. These flowgraphs interconnect
different signal processing modules via a vertex which originates from a source
module and terminates at a sink module, thus representing the flow of the data. The
application flowgraph can be written in either C++ language or python language,
while the core signal processing modules are written in C++ language. The GNU
Radio comes with a graphical UI and called as GNU Radio Companion. The GNU
Radio Companion provides drag-n-drop functionality to configure a particular appli-
cation, thus allowing a novice user without having any programming knowledge to
utilize this toolkit. GNU Radio is natively available for Linux, but other alternate
build methods are available for Windows and MAC OS.
2.4.3.2 Pothosware
Pothosware is an open source software for interconnected processing blocks [29]. It

provides multiple toolkits for different purposes. Pothos provide toolkit for interfac-
ing with SDR hardware, provides another toolkit for real-time signal analysis and
multiple toolkits signal processing blocks. Pothos also allows to integrate different
components from other SDR softwares into the pothosware. The Pothos SDR toolkit
provides a SDR source block for receivers and SDR sink block for transmitters. For
interfacing with different hardwares, the Pothos project wraps SoapySDR project.
SoapySDR is an open source C/C++ library for interfacing with different types of
SDR hardware. The library is vendor neutral, implying that it is compatible with
different types of hardware.
There exist multitude of software defined radio platforms, including Universal
Software Radio Peripheral (USRP), eFalcon, Microsoft Research Software Radio
(Sora), Cognitive Baseband Reconfigurable Radio (COBRA), Berkeley Emulation
Engine (BEE2), and Wireless open-Access Research Platform (WARP). To keep it
short and interesting, we will discuss one platform from traditional SDR: WARPnet
and one unconventional SDR hacked platform: RTL-SDR.
2.4.4 WARPnet
The platform is build from ground up by integrating individual blocks at the lower
layer to realize a whole cognitive radio functional platform [30]. The platform allows
the developers to fine control the lower layer functionalities while allowing conven-
tional users to deploy the platform for field testing. It utilizes two wide band radio
boards to support desired transmission characteristics. One is MAX2829 which is
dual band 2.4 and 5 GHz radio, while the other radio utilized is AD9352 which is
a 2.5 GHz band radio. The radio board utilizes multiple dedicated control interfaces
(channels) for coordination between different boards.
The WARPnet board utilizes a Xilinix Virtex-4 FX series FPGA. This FPGA
provides the processing power required for doing complex computation over software
for the working on real-time algorithms for wireless communication. The board has
Marvell Alaska 88e1111 Gigabit Ethernet transceiver for bridging wireless to wired
network infrastructure. The board is capable of supporting 2 GB of RAM. The board
also has two 6.5 Gbs SATA bus for connecting hard drives for data storage. The
board also support other hardware features which are not discussed in the book.
Readers are suggested to go through the WARPnet documentation to get the detailed
specifications.
Maxim MAX2829 RF transceiver provide analog baseband interfaces and handles
RF translations. The transmitter has 16 bit dual input Analog Devices’ AD9777 DAC.
The receiver has dual 14bit output AD9248 ADC. The transceiver provide software
control over transmit power, input and output filter etc. via SPI interface.
Analog device’s AD9352 transceiver chip integrates the ADC and DAC along
with the RF transceiver, filters and amplifiers. A dedicated Spartan-3 AN FPGA is
utilized for providing control and data interface along with the processing algorithms
for radio.
The board also utilizes a backdoor board which provides dedicated control channel
and a control hub. This allows as a central controller to control all boards in the
network. This includes remote programming of deployed boards in the network.
The backdoor interface utilizes Ethernet and 900 MHz long range wireless interface
for communication. The board utilizes GPS module to synchronize time with other
boards for the purpose of MAC layer algorithms.
WARPnet architecture utilizes three functional cores: FPGA logic, the PowerPC
and the Axis Etrax SoC. The high sped FPGA implements the physical layer of
the cognitive radio via utilization of complex designs in HDL. The MAC layer is
implemented in C/C++ language via the PowerPC as it has direct access to the lower
physical layer. The overall cognitive control of the WARPnet node is done via the
Axis Etrax SoC which provide control channel and monitoring of the network. In a
WARPnet network, each node is connected and managed by a central entity called
as WARPnet controller. The WARPnet controller monitors the system characteris-

tics and provides a system wide view of the whole deployment. This allows easy
calculation of the performance of the grid.
2.4.5 RTL-SDR
RTL-SDR is a software defined radio that is commonly found in the DVB-TV

tuner dongle which utilizes Realtek RTL2832U chipset [31]. It was found that the
RTL2832U chipset allows transferring raw I/Q samples to the host, thus allowing
demodulation to be done in the software. The DVB-TV tuner is a cheap dongle
which acts as a receiver, therefore with proper tools it can be utilized as a SDR
receiver. The dongle utilizes a combination of a tuner, sampler and USB interface
modules. The RTL2832U supports tuners at Intermediate Frequency of 36.125 MHz,
low-Intermediate Frequency of 4.57 MHz, or Zero-Intermediate Frequency.
The RTL2832U provides 8 bit ADC sample with maximum theoretical sampling
rate of 3.2 MS/s along with USB interface for pumping data to the host machine.
Although the maximum bandwidth is 3.2 MHz, there are sample losses. The largest
sample rate successfully without sample loss is 2.8 MHz. Different tuners provide
different types of signal reception in different range of frequencies. Table 2.5 provides
the list of tuners which is usually utilized with RTL2832U [32].
Although this hardware does not provide any transmission capability, it provides
sensing capability which can be utilized as a starting point for people entering into
cognitive radio field. The hardware does not provide accuracy to industry standards,
but with proper noise/EMI protection, heat sinks and stable oscillator, a high level
of accuracy can be obtained. The librtlsdr library provides generic driver interface
for RTL2832U chipset [32]. Therefore most of the software defined radio software
can be interfaced with this dongle via the help of librtlsdr library. The applications
include FM radio, GSM listening, ADSB receiver, GPS receiver, AIS receiver etc.
Table 2.5 Different tuners Tuner Frequency band

with supported frequency
bands Rafael Micro R820T 24–1766 MHz
Rafael Micro R828D 24–1766 MHz
Fitipower FC0013 22–1100 MHz
Fitipower FC0012 22–948.6 MHz
FCI FC2580 146–308 MHz and
438–924 MHz (gap in
between)
Elonics E4000 52–2200 MHz with a gap
from 1100 to 1250 MHz
2.5 Discussion
In this chapter various cognitive radio network technologies for QoS provisioning
and energy management related works were reviewed and background materials
were described which are essential to appreciate the following chapters. First, various
spectrum management strategies were discussed and requirements of media access
control techniques were introduced. Towards the end of the chapter, various energy
management techniques were also discussed. Further a brief overview on different
cognitive radio platforms was also described. The capabilities of these platform
indicates that realization of the cognitive radio technologies. In the following chapters
we focus exclusively on a framework which provides QoS provisioning and energy
management for cognitive radio network, encompassing the technologies of spectrum
management, media access scheme, energy management and also self-coexistence.
References
1. FCC. (2003). Revision of parts 2 and 15 of the commissions rules to permit unlicensed national
information infrastructure (U-NII) devices in the 5 GHz band. Memorandum opinion and order
(Vol. 21).
2. Atakan, B., & Akan, O. B. (2007). Biologically-inspired spectrum sharing in cognitive radio
networks. In 2007 IEEE Wireless Communications and Networking Conference (pp. 43–48).
3. Eletreby, R. M., Elsayed, H. M., & Khairy, M. M. (2014). Optimal spectrum assignment for
cognitive radio sensor networks under coverage constraint. IET Communications, 8(18), 3318–
3325.
4. Niyato, D., Member, S., Hossain, E., & Member, S. (2008). Competitive pricing for spectrum
sharing in cognitive radio networks: Dynamic game, inefficiency of nash equilibrium, and
collusion. IEEE JSAC, 26(1), 192–202.
5. Song, Y., Fang, Y., & Zhang, Y. (2007, November). Stochastic channel selection in cognitive
radio networks. In Global Telecommunications Conference, 2007. GLOBECOM ’07. IEEE
(pp. 4878–4882).
6. Talat, S., & Wang, L.-C. (2009). Qos-guaranteed channel selection scheme for cognitive radio
networks with variable channel bandwidths. In International Conference on Communications,
Circuits and Systems, 2009. ICCCAS 2009 (pp. 241–245).
7. Cai, L. X., Liu, Y., Shen, X., Mark, J. W., & Zhao, D. (2010). Distributed QoS-aware MAC
for multimedia over cognitive radio networks. In Global Telecommunications Conference
(GLOBECOM 2010), 2010 IEEE (pp. 1–5).
8. Wang, P., Niyato, D., & Jiang, H. (2010). Voice-service capacity analysis for cognitive radio
networks. IEEE Transactions on Vehicular Technology, 59, 1779–1790.
9. Timmers, M., Pollin, S., Dejonghe, A., Van der Perre, L., & Catthoor, F. (2010). A distributed
multichannel MAC protocol for multihop cognitive radio networks. IEEE Transactions on
10. Cordeiro, C., & Challapali, K. (2007). C-MAC: A cognitive MAC protocol for multi-channel
wireless networks. In 2nd IEEE International Symposium on New Frontiers in Dynamic Spec-
trum Access Networks, 2007. DySPAN 2007 (pp. 147–157).
11. Zhao, Y., Song, M., Xin, C. (2013). FMAC: A fair MAC protocol for coexisting cognitive radio
networks. In INFOCOM, 2013 Proceedings IEEE (pp. 1474–1482).
12. IEEE 802.11 Standard (2012). IEEE 802.11 working group, wireless LAN medium access
control (MAC) and physical layer (PHY) specifications.
References 95
13. Le, L., Hossain, E. (2008). A MAC protocol for opportunistic spectrum access in cognitive
radio networks. In Wireless Communications and Networking Conference, 2008. WCNC 2008.
IEEE (pp. 1426–1430).
14. Mishra, V., Tong, L. C., Chan, S., & Mathew, J. (2011). MAC protocol for two level QoS
support in cognitive radio network. In 2011 International Symposium on Electronic System
Design (ISED) (pp. 296–301).
15. Wang, S., Wang, Y., Coon, J. P., & Doufexi, A. (2012). Energy-efficient spectrum sensing and
access for cognitive radio networks. IEEE Transactions on Vehicular Technology, 61, 906–912.
16. Maleki, S., Pandharipande, A., & Leus, G. (2011). Energy-efficient distributed spectrum sensing
for cognitive sensor networks. IEEE Sensors Journal, 11, 565–573.
17. Liu, Y., Xie, S., Zhang, Y., Yu, R., & Leung, V. C. M. (2012). Energy-efficient spectrum
discovery for cognitive radio green networks. Mobile Networks and Applications, 17(1), 64–
74.
18. Gao, S., Qian, L., & Vaman, D. R. (2008). Distributed energy efficient spectrum access in wire-
less cognitive radio sensor networks. In 2008 IEEE Wireless Communications and Networking
Conference (pp. 1442–1447).
19. Zhang, H., Zhang, Z., Chen, X., & Yin, R. (2011). Energy efficient joint source and chan-
nel sensing in cognitive radio sensor networks. In 2011 IEEE International Conference on
Communications (ICC) (pp. 1–6).
20. Zimmermann, A., Günes, M., Wenig, M., Ritzerfeld, J., & Meis, U. (2006). Architecture of
the hybrid MCG-mesh testbed. In Proceedings of the 1st International Workshop on Wireless
Network Testbeds, Experimental Evaluation and Characterization, WiNTECH ’06 (pp. 88–89).
New York: ACM.
21. Tan, K., Wu, D., Chan, A., & Mohapatra, P. (2010). Comparing simulation tools and exper-
imental testbeds for wireless mesh networks. In 2010 IEEE International Symposium on a
World of Wireless Mobile and Multimedia Networks (WoWMoM) (pp. 1–9).
22. Thompson, J. J., Hopkinson, K. M., & Silvius, M. D. (2015). A test methodology for evaluating
cognitive radio systems. IEEE Transactions on Wireless Communications, 14, 6311–6324.
23. Masonta, M. T., Mzyece, M., & Mekuria, F. (2012). A comparative study of cognitive radio
platforms. In Proceedings of the International Conference on Management of Emergent Digital
EcoSystems, MEDES ’12 (pp. 145–149). New York: ACM.
24. Gustafsson, O., Amiri, K., Andersson, D., Blad, A., Bonnet, C., Cavallaro, J. R et al. (2010)
Architectures for cognitive radio testbeds and demonstrators 2014; an overview. In 2010 Pro-
ceedings of the Fifth International Conference on Cognitive Radio Oriented Wireless Networks
and Communications (pp. 1–6).
25. Lotze, J., Fahmy, S. A., Noguera, J., Doyle, L., & Esser, R. (2008). An FPGA-based cognitive
radio framework. In Signals and Systems Conference, 2008. (ISSC 2008). IET Irish (pp. 138–
143).
26. Mitola, J., III, & Maguire, G.Q., Jr. (1999). Cognitive radio: Making software radios more
personal. IEEE Personal Communications, 6, 13–18.
27. Kuon, I., Tessier, R., & Rose, J. (2008). FPGA architecture: Survey and challenges. Foundations
and Trends in Electronic Design Automation, 2, 135–253.
28. Blossom, E. (2004). Gnu radio: Tools for exploring the radio frequency spectrum. Linux Jour-
nal, 2004, 4.
29. SoapySDR Pothosware. (2016) Retrieved April 04, 2016, from http://www.pothosware.com.
30. Gupta, S., Hunter, C., Murphy, P., & Sabharwal, A. (2009). Warpnet: Clean slate research on
deployed wireless networks. In Proceedings of the Tenth ACM International Symposium on
Mobile Ad Hoc Networking and Computing, MobiHoc ’09 (pp. 331–332). New York: ACM.
31. Wickert, M. A., & Lovejoy, M. R. (2015) Hands-on software defined radio experiments with the
low-cost RTL-SDR dongle. In Signal Processing and Signal Processing Education Workshop
(SP/SPE), 2015 IEEE (pp. 65–70).
32. RTL-SDR. Retrieved November 01, 2016, from http://www.rtl-sdr.com/about-rtl-sdr.
Chapter 3
QoS Provisioning and Energy Management
Framework for CRN
The cognitive radio comprises of the following three important functional compo-
nents to realize its operation as wireless communication device:
1. Sensing: Sensing is at the center of the cognitive radio platform. This is responsible
for sensing the environment and providing opportunistic information to the cognitive
radio platform. Sensing information can be utilized locally or it can be passed to other
devices responsible for utilizing the spectrum information. Examples would include
passing the information to peer cognitive radio device for distributed spectrum usage
learning. Another example would be passing the information to a central server which
is responsible for making decisions.
2. Acclimation: A cognitive radio based on the sensed environment adapts its oper-
ating parameters to suit the desired operating behavior. The operating parameters
primarily includes PHY and MAC layer of a cognitive radio device, but other OSI
layers may need adjustment depending upon the network behavior.
3. Transmission/Reception: Once the transmission parameters are adapted, the cog-
nitive radio node communicates with its peer node. To achieve such communication
every node in the cognitive radio system need to adhere to a common standard of
operation. This is required for any communication system to function properly.
Since cognitive radio operates in a sporadic heterogeneous environment, QoS
provisioning in such system faces various challenges. A framework for cognitive
radio network is discussed which includes functional modules responsible for QoS
provisioning. To achieve this, an observation over the macroscopic and microscopic
behavior of resource demand including bandwidth and latency is required. Different
frameworks have been proposed and implemented by researchers for cognitive radio
networks [1–4]; However not many framework focuses on the issue of QoS provi-

98 3 QoS Provisioning and Energy Management Framework for CRN
sioning in cognitive radio network from the perspective of Layer 2 of the protocol
stack.1 This book will integrate different technologies responsible for QoS and there
relationship with each other.
3.1 QoS Parameters
As described in previous section, quality of service is defined as the ability of

a system to provide a guarantee over a minimum grade of service offered to an
inbound/outbound traffic. The two important parameters of QoS are latency and
throughput. These two parameters have correlation with respect to resource access
[5]. The system has varying degree of latency if the channel access is correlated with
the bandwidth resource.
3.1.1 Latency Versus Throughput
Multiple factors affect latency including propagation delay, serialization delay, rout-
ing/switching delay, queue and buffer management delay [6]. For a user with band-
width requirement more than the resource available, the latency is caused by an
additional parameter called as resource access delay. A resource access delay of a
packet refers to the time period for which the packet was deprived of resource since
it was next in line for transmission in the queue [7]. This type of delay depends
upon the access scheme used and hence the QoS provisioning framework includes
resource access scheme as one of its important ingredient component. Throughput on
the other hand refers to the rate of successful message delivery over a media [8]. This
is affected by the bandwidth and quality of the selected wireless channel and also on
the access scheme utilized by the user. Therefore channel selection is also included
as one of the important ingredient composition of QoS provisioning framework.
To realize the quality of service, packet categorization is done to implement access
categories. As discussed in Sect. 1.5, four access categories are supported by the
discussed framework as shown in Table 3.1. Voice data is provided highest priority
while background data is provided lowest priority. Therefore any packet generated
by any application will fall within one of these categories. These access categories
have different priorities and hence different handling techniques. Packets from higher
access category are not dropped as often as packets from lower access category. Since
a user may run multiple applications each generating traffic of different priority,
1 Layers
of the protocol stack refers to the architectural layers of the Open System Interconnection
(OSI) model. This book will utilize the term protocol stack to represent functional layers of the OSI
model.
3.1 QoS Parameters 99
Table 3.1 Priorities of different packets

Access category Description Marker
Voice priority Highest priority, Low latency 0
Video priority Second highest priority, Video 1
can buffer if needed
Best effort priority If no QoS mentioned, or burst 2
mode traffic like web surfing
Background priority Lowest priority, print jobs, no 3
strict latency
packets are identified using packet marking scheme in packet headers [9]. Through
the marking of packets in the header, the protocols realize which packet belongs to
which category.
3.1.2 Self-coexistence and Its Role in QoS
Since wireless channels are unguided media (except in the case of beamforming,
which is not considered in this work), operation of multiple wireless networks in
the vicinity of each other suffers from interference [10]. Self coexistence refers to
the method of establishing a non-interfering coexistence with other homogeneous
networks in the vicinity. To achieve this coexistence, multiple methods have been
proposed and most of them require sacrificing a part of resource so that interference
is minimized. These sacrifices result in deterioration of QoS in the network. There-
fore a method is required to cater graceful degradation of QoS while mitigating the
interference for self-coexistence. This yields in the composition of self-coexistence
management functionality in the QoS provisioning framework for cognitive radio
network.
3.1.3 Energy Management and QoS
For green communication and to prolong the life of a battery operated device, it
is fundamental to reduce the overall power consumption of the wireless network.
This is achieved by reducing the transmission power and by operating on channels
which require lower transmission power. Reducing the transmission power reduces
the SNR/SINR of the signal and hence the QoS. Similarly, operating on channels with
lower transmission power with lower bandwidth results in bandwidth loss and hence
a degradation in QoS. Therefore an energy management system is incorporated as an
ingredient of the composition of framework for the purpose of graceful degradation
of QoS.
3.2 QoS Framework for Cognitive Radio Network
The high level view of the discussed QoS framework at the Layer 2 of the protocol
stack is shown in Fig. 3.1. The figure shows different blocks/units along with their
inter-network connections depicting the relationship between the components and
the type of interaction.
The discussed framework works alongside the normal communication path allow-
ing the re-utilization of conventional protocol stack. The application domain consists
of multiple applications including the applications requiring QoS support. These
applications could be multimedia application, VoIP, gaming etc. This domain con-
veys the QoS requirements of the applications to the other units in the framework.
QoS control unit (QCU) is the block which receives the information from the appli-
cation domain regarding the requirement of the QoS. It conveys this information to
the Channel selection and management unit (CSMU) as well as to the MAC protocol
unit (MPU). Environmental and RF channel block provide the information about the
external environment including channel availability, and this information is passed
to the CSMU and also to the MPU. The Energy Management Unit (EMU) is used
to make important decisions with the help of other units to conserve the energy of
the user. Channel Selection and Management Unit (CSMU) based on the knowledge
Application Domain
Channel
Energy Selection Self
QoS Management and Coexistence
Control Unit Unit Management Engine
Unit
MAC Protocol Unit
Environment and RF Channel
Fig. 3.1 QoS provisioning framework for cognitive radio network

3.2 QoS Framework for Cognitive Radio Network 101
from QCU select the channel which meets the required QoS level. Also, if required
the CSMU utilizes the information from EMU to select a channel which will conserve
the energy of the system. MAC Protocol Unit (MPU) considers all the information
from all other units and selects/design appropriate MAC which conforms to all the
requirements. If all requirements are not met, then an optimization step is required
to achieve the desired level of result. A self-coexistence engine (SCE) is required to
mitigate the issues of mutual interference experienced by multiple cognitive radio
networks operating in the vicinity of each other. This unit interacts with CSMU and
MPU to provide self-coexistence from the perspective of resource allocation and
resource sharing respectively.
3.3 Detailed Layer 2 QoS Provisioning Framework
Figure. 3.2 shows the detailed view of the discussed framework for QoS provisioning
at the Layer 2 of the protocol stack. The right portion shows the normal communi-
cation system path, in which the application sends the packets to the routing engine
and from there it is sent to the MAC instance. A similar process happens in the
reverse direction whenever a packet is received by the MAC instance. The left por-
tion shows the QoS engine which works in parallel with the normal communication
path. This portion receives the QoS requirement from the application (if application
can provide) and performs actions needed to achieve the desired QoS with the help
of different functional modules.
The framework is composed of different individual functional modules or com-
ponents, which collectively operate towards achieving the common goal of network
operation. The complete framework is interrupt driven and therefore any module
would perform busy waiting until it receives an interrupt from source module. These
modules are connected via function call and response method; parameters are passed
via function calls and results are obtained in the form of response. Following are the
description of different modules:
Application: Application is the user process generating traffic with varying priori-
ties. User applications are assumed to generate 4 levels of priority traffic as shown
in Table 3.1. Since a user can support multiple applications with varying levels of
priorities, the application process allows multiplexing of different data types. An
application can also act as a process for generating system traffic for the purpose of
control plane.
Application QoS Engine: This component converts the application QoS requirement
into the discussed framework language. For example if the application identifies
the traffic communicated as VoIP traffic, application QoS engine will convert this
information into the QoS bound value of the VoIP traffic. The QoS bounds are defined
in terms of latency and bandwidth; and these values are stored in a database which
are accessible to network operator. This component is integrated within the QoS
Control Unit.
QoS Engine Communication

stack
Application
Application QoS
Engine
Packet
Routing Module
Classification Unit
Energy Management
QoS Control Unit Packet Admission
Unit
Control
Self
Coexistence
Engine
Channel Selection and

Management Unit Policy
Database domain
MAC instance
MAC Protocol Unit
MAC instance
Layer 2
and above
Layer 1
Success/Failure
report
Physical Module / RF Channel
Sensing report
Fig. 3.2 Detailed QoS provisioning framework for cognitive radio network
QoS Control Unit: This unit is responsible for procuring and facilitating the QoS
requirements of different traffic. Any other module/component may demand for QoS
bound for the purpose of its operation. This unit is also responsible for managing
and monitoring the experienced QoS of the system. QCU monitors the QoS of the
system which it experiences through the utilization of current mode of the protocol
stack with its current parameters. A feedback based control is utilized to manage the
QoS by replacing/configuring the appropriate module which in turn will guide the
QoS of the system between the defined bounds.
3.3 Detailed Layer 2 QoS Provisioning Framework 103
Policy Domain: Policy domain provides two sets of radio regulation policies: local
and global, for a given geographical location. The local policy domain always super-
sedes the global policy domain. In case there is no entry in local policy domain for
the given geographical location, global policy domain is accounted in. These radio
regulation policies include allowed channels, maximum allowed power, operating
bandwidth, channel separation, maximum usage time, sensing duration and sensitiv-
ity. Apart from providing these regulation policies, policy domain may also provide
a list of optimal channels for selection. These optimal channels may be provided by
third party spectrum monitoring institutes.
Channel Selection and Management Unit: Normally, channel selection is handled
by PHY module (Layer 1) of a communication system, but in the discussed frame-
work, this functionality is included at Layer 2 so as to enable close integration with
other QoS related functionalities. The channels from Layer 1 are abstracted as iden-
tifiers to the Layer 2 along with their characteristics including bandwidth, occupancy
statistics, SINR and transmission power requirement. Channel selection and man-
agement unit (CSMU) is responsible for selection of single or multiple wireless
channels from a given list of channels. The channels are selected and facilitated
to the media access scheme for communication purposes and therefore the number
of channels selected depends upon the media access scheme. The given list from
which the CSMU has to select channel is provided by policy domain. Various fac-
tors affect the selection of channel including occupancy, QoS constraint (latency and
bandwidth) and energy consumption bound.
Energy Management Unit: Energy management unit provides the information about
the remaining energy in the system. This unit helps in deciding whether a battery
saving protocol should be activated or not. If a battery saving protocol is activated
it provides information to the CSMU about the energy consumption bounds for the
purpose of channel selection. In this way an optimized protocol can be used so as
to minimize the energy consumption. Since the system tries to minimize the energy
consumption, a graceful degradation in QoS is achieved with the help of packet
admission control administered by energy management unit.
Packet Admission Control: Packet admission control unit is responsible for block-
ing/dropping certain number/category of data packets so as to provide resources
to other higher priority data packets. This is mainly done when the total resource
demand is greater than the available resource. Low priority packets are blocked in
the queue and are transmitted when the resource available is commensurate to the
demand. If resources are not available for prolonged time, the packets are dropped
from the queue.
Traffic Classification Unit: Legacy applications may not have the functionality to
explicitly provide the information about the priority of a traffic; this is overcome
by providing a traffic classification unit. The work of traffic classification unit is
to classify the incoming/outgoing traffic into one of the mentioned access category
(Table 3.1). Different traffic classifiers utilize different methods for building up a
classifier including statistical and rule based learning. Study of traffic classifier is
beyond the scope of this work.
Routing Engine: Routing engine is a layer 3 technology which provides opportunistic
routing protocols for packet routing. This unit is responsible for efficient routing,
congestion control and energy efficient data transfer across the network. Routing
engine may utilize information from other modules presented in this book. Since it
is a layer 3 technology, it is beyond the scope of this book.
MAC Protocol Unit: This unit spawns/selects a MAC protocol which accommodates
the requirements of QoS bounds of the given application. This unit gets the informa-
tion about the QoS requirements from QCU. On receiving this information, the MPU
retrieves the list of channels from CSMU and utilizes it to spawn a MAC instance.
This unit is capable of generating multiple MAC instances for multi-radio system.
In case of multi-radio system, all MAC instances operate in parallel by utilizing a
multiplexing and demultiplexing scheme. This work does not consider the case of
multiple radio.
Environment and RF Channel: This Layer 1 unit is equipped with radio which
performs sensing and communication. Therefore reconfiguration of mixers, filters,
modulator/demodulator and detector is done in this unit. Apart from reconfigura-
tion of physical communication layer, this unit provides information about channel
conditions like noise power, signal power, bit error rate etc. Since, this is a layer 1
technology, it is beyond the scope of this work.
Sensing Report: This unit encapsulates a spectrum detection functional module to
provide the information about the availability of a channel. It receives informa-
tion about channel parameters from Environment and RF module and facilitates
CSMU for enhanced channel selection whereas the same information is utilized
by an instance of MAC protocol to utilize/skip the frame. Sensing report unit is
abstracted so as to present only the relevant information to the upper layers.
Success/Failure Report: This unit processes the information about the success/failure
rate of the packets from the Environment and RF channel unit and pass it to the
Channel Selection and management unit, so as to improve the decision on channel
selections. The same information is passed to the MAC protocol unit which utilizes
this information and if necessary, reconfigures its MAC instance on the fly. Again,
this unit is abstracted so as to present only the relevant information to the upper
layers.
3.3.1 Mode of Operation
Since cognitive radio is based on the technology of reconfiguration of protocols and

communication parameters according to the environment in which it operates, the
system operates with a particular set of parameters and protocols at a given time. This
3.3 Detailed Layer 2 QoS Provisioning Framework 105
particular instance of protocol stack is termed as mode of protocol stack. In the dis-
cussed framework, a mode of protocol stack comprises of: Network Mode, Spectrum
Selection Mode, Media Access Mode and Self Coexistence Mode. The protocol stack
in such system is termed as multi-mode protocol stack. A network mode conveys
the information about the mode of operation of the network: infrastructure or ad-
hoc. Based on this mode the MPU unit reconfigures its selected media access mode
for network operation. Spectrum selection mode provides the information about the
spectrum selection technique (protocol) that is being utilized by the protocol stack.
Similarly, self-coexistence mode provide information about the type self-coexistence
protocol utilized by the protocol stack.
3.3.2 Generic Protocol Stack
The discussed framework utilizes different modules which share common function-
alities and therefore a generic protocol stack can be introduced in control plane to
build an efficient multi-mode capable system. To realize a generic protocol stack, the
common functionalities are brought together to form a toolbox of protocol [11]. A
generic protocol stack allows code/module sharing and accelerated protocol devel-
opment using code re-usability. Therefore generic protocol stack has low memory
footprint in comparison to other multi-mode protocol stack enabled by composition
of different single mode protocol stacks.
Figure 3.3 shows the UML class diagram of a reconfigurable generic multi-mode
protocol stack. As seen from the diagram, a generic multi-mode protocol stack com-
reconfigurable (multi-
single protocol mode) protocol stack
stack
system specific reconfiguration function/

protocol stack management plane
specific functionality common functionality
Specific modules generic protocol stack
mode 1 mode 2 protocol functions protocol architecture
mode n
data structures
Fig. 3.3 UML diagram of the generic multi-mode protocol stack

prises of various elements that can be identified as analogous to each other in the
context of communication protocols. Some of these commonalities include:
1. Functionality provided by a certain layer. An example would be parametrized
function for selecting optional RTS/CTS mechanism in MAC layer.
2. Protocol architecture and framework, for instance structure of a MAC frame
(sequence, duration etc.) and access rules may be shared across different MAC
schemes.
3. Data structure or protocol data unit (PDU) utilized by a protocol can be utilized
in a generic way across layer.
These common functionalities together with specific functionalities implements a
system specific protocol stack. A reconfigurable multi-mode protocol stack is real-
ized with the help of a reconfiguration function which parametrizes the generic and
mode-specific modules. Reconfiguration function is realized via QoS control unit
in management plane, which guarantees the convergence of fused complementary
generic and mode-specific modules for spawning a system specific protocol stack.
3.3.2.1 Generic Versus Specific Partition
The partitioning of generic and specific functionality (system specific protocol stack
or proprietary stack or specific modules) of a particular protocol is beyond the scope
of this book. Nevertheless, a brief idea about the partitioning of functionalities is
provided. In software engineering, a generic component refers to the identifica-
tion of requirements that are met by a large family of abstraction but restrictive
enough to efficiently realize its function. Many functions utilized by protocol stack
are mode-independent, i.e. they are shared across all modes of operation. Some
generic functions can return mode-specific results by providing appropriate mode-
specific parameters in a function call [12]. An example would be implementation of
automatic repeat request (ARQ) protocol in link layer, which can be configured by
a parameter to behave as Go-back-N or Selective Reject ARQ.
In comparison to generic component, a mode-specific component refers to the
function which is unique to a mode and cannot be implemented on shared basis with
other modes. An example would be the implementation of Hybrid ARQ which does
not share any commonality with traditional ARQ mechanism. These mode-specific
components are complemented by their counterpart generic component to realize
a protocol layer. Therefore each layer is composed of a generic and mode-specific
part. A slice of particular instance of protocol stack with generic and mode-specific
component in each layer is shown in Fig. 3.4.
3.4 Self Organization 107
Reconfiguration plane
Management plane
User and control plane Reconfiguration

QoS Control Unit
RLC-g RLC-g
Service access RLC-s1 RLC-s2

point
Self organization
MAC-g MAC-g
MAC-s1 MAC-s2
PHY-g
PHY-g
PHY-s1 PHY-s2
Protocol stack in Mode 1 Protocol stack in Mode 2
Fig. 3.4 Self organization of the protocol stacks between different modes of operation
3.4 Self Organization
The discussed framework can be configured to manually/automatically select a

desired mode of operation of protocol stack. The Network mode is manually selected
to operate a network in infrastructure or ad-hoc mode. All other modes can be con-
figured to automatically select a desired mode of operation of the protocol stack
so as to guarantee a minimum grade of quality of service. The self organization is
done in management plane by QoS control unit. Therefore QCU continuously mon-
itors the system overall performance in terms of latency, packet loss and achievable
throughput.
Figure 3.4 shows a slice of protocol stack reconfiguring from mode 1 to mode 2
and vice-versa. The protocol reconfiguration happens at the reconfiguration plane
which is controlled and managed by the management plane. Management plane
functionality is realized by QoS control unit which continuously monitors the QoS
of the system and reconfigures the protocol stack accordingly. The generic and mode-
specific components are defined in user and control plane. Each layer of the protocol
stack is connected to other layer via a service access point. Service access point acts
as the interface for transfer of protocol data unit across layers. PHY-s1, MAC-s1 and
RLC-s1 (Radio Link Control) represent the mode 1 specific components of physical,
media access control and radio link control protocol respectively. Similarly PHY-
g, MAC-g and RLC-g represent the generic component of physical, media access
control and radio link control protocol respectively. These generic components are
re-utilized in both modes as shown in Fig. 3.4. A self organization between different
modes of operation can be initiated by management plane to achieve a certain grade
of QoS in the system.
3.4.1 Channel Availability Model
The frequency bands utilized in the system are constituted as channels identified by
an ID or a channel number. These channels can be homogeneous or heterogeneous
depending upon the system model utilized for simulation purposes. Since modelling a
licensed user traffic pattern on the channel is tedious in nature, the channel availability
is modelled instead. To do so, each channel is assumed as an ON-OFF source. When a
channel is ON, it implies that channel is occupied by licensed user and is unavailable
for secondary purposes. Similarly, when a channel is OFF, it implies that channel is
unoccupied and a secondary user can utilize the channel. Therefore a channel can be
modelled as a two-state Markov chain as shown in Fig. 3.5. The transition probability
α represents the transition probability from ON-state of the channel to its OFF-state.
Similarly, the transition probability β represents the transition probability from OFF-
state of the channel to its ON-state. The steady state probability of a channel to be
free from incumbent activity is calculated as:
αi
i = ∀i ∈ {1, ..., C} (3.1)
αi + βi
Similarly, the steady state probability of a channel to be occupied by incumbent is

calculated as:
βi
ρi = ∀i ∈ {1, ..., C} (3.2)
αi + βi
3.4.2 Quiet Periods
Standards designed by IEEE 802.22 WRAN group suggest that the maximum inter-
ference duration that a licensed user operating on a TV band can tolerate is 2 s [13].
Therefore according to the IEEE 802.22 standards, a cognitive access point should
perform channel sensing every 2 s so as to ensure the protection of licensed incum-
bents. In the discussed framework, channels are sensed every 100ms. While sensing
channels, it is made sure that all nodes (CR users) stop their communication for
Fig. 3.5 Two state Markov

chain availability model of a
channel. ON state represent
that channel is unavailable
and vice-versa
3.4 Self Organization 109
the duration of sensing period. This duration is termed as quiet period. The design
of quiet period follows the design of media access scheme. There are two types of
sensing performed by a node: in-band and out-of-band.
3.4.2.1 In-Band Sensing
In-band sensing refers to the sensing performed by nodes on the channel in which
they currently operate on. Since nodes normally exchange information with other
nodes operating on the same channel, the distribution of quiet period for in-band
sensing is easily obtained.
3.4.2.2 Out-of-band Sensing
Out-of-band sensing refers to the sensing performed by nodes in channels other

than on which they are currently operating. Since sensing is performed in different
channel, the distribution of quiet period on that channel needs to be obtained by the
node performing out-of-band sensing. The design of information interchange about
quiet period across all nodes in the network operating on different channels need to
be considered while designing a media access scheme.
3.5 Discussion
Most of the research on QoS provisioning for CRN has been done individually
without introducing a formal framework. Since cognitive radio is based on the tech-
nology of reconfiguration, a standard framework for CRN will allow easy and formal
integration of different technologies responsible for QoS. The identification and com-
partmentalization of different technologies responsible for QoS will allow abstract
and encapsulated behavior of framework design.
To achieve the above, a QoS provisioning framework for cognitive radio network
at Layer 2 of the protocol stack is utilized. Different functional modules are identified
for the purpose of QoS provisioning from the perspective of latency and throughput.
While latency and throughput can be directly identified by the media access scheme
and channel selection technique respectively, the effect of self-coexistence and energy
management is separately studied. Macroscopic and microscopic functional modules
at layer 2 of the protocol stack are discussed to provide an in-depth idea about the
requirements. The concept of generic and mode-specific protocol stack is introduced
to envisage the building blocks of the framework. From implementation and operation
point of view, self-organization technique of multi-mode stack is also discussed.
References
1. Rondeau, T., & Bostian, C. (2009). Artificial intelligence in wireless communications. Artech
House mobile communications series. Norwood: Artech House.
2. Ansari, J., Zhang, X., Achtzehn, A., Petrova, M., & Mahonen, P. (2011). A flexible mac devel-
opment framework for cognitive radio systems. In Wireless Communications and Networking
Conference (WCNC), 2011 IEEE (pp. 156–161).
3. Wysocki, T., & Jamalipour, A. (2009). Mac framework for intermittently connected cognitive
radio networks. In 2009 IEEE 20th International Symposium on Personal, Indoor and Mobile
Radio Communications (pp. 481–485).
4. Rieser, C., Rondeau, T., Bostian, C., Cyre, W., & Gallagher, T. (2006). Cognitive radio engine
based on genetic algorithms in a network. Jan 12 2006, US Patent App. 10/875,619.
5. Gamal, A., Mammen, J., Prabhakar, B., & Shah, D. (2004). Throughput-delay trade-off in
wireless networks. In INFOCOM 2004. Twenty-third AnnualJoint Conference of the IEEE
Computer and Communications Societies (Vol. 1, p. 475).
6. Doudou, M., Djenouri, D., & Badache, N. (2013). Survey on latency issues of asynchronous
mac protocols in delay-sensitive wireless sensor networks. Communications Surveys Tutorials,
IEEE, 15, 528–550.
7. Sakurai, T., & Vu, H. (2007). MAC access delay of IEEE 802.11 DCF. IEEE Transactions on
Wireless Communications, 6, 1702–1710.
8. Bianchi, G. (1998). IEEE 802.11-saturation throughput analysis. IEEE Communications Let-
ters, 2, 318–320.
9. Liu, C. (2007). Tracking DSCP marking of packets in a QOS enabled triple-play IP network.
In Third International Conference on Networking and Services, 2007. ICNS (pp. 21–21).
10. Gardellin, V., Das, S., & Lenzini, L. (2013). Self-coexistence in cellular cognitive radio net-
works based on the IEEE 802.22 standard. IEEE Wireless Communications, 20, 52–59.
11. Sachs, J. (2003). A generic link layer for future generation wireless networking. In IEEE
International Conference on Communications, 2003. ICC ’03. (Vol. 2, pp. 834–838).
12. Siebert, M., & Walke, B. (2001). Design of generic and adaptive protocol software (DGAPS).
In Proceedings of 3G Wireless (Vol. 1).
13. Cordeiro, C., Challapali, K., Birru, D., & Sai Shankar, N. (2005). IEEE 802.22: The first world-
wide wireless standard based on cognitive radios. In 2005 First IEEE International Symposium
on New Frontiers in Dynamic Spectrum Access Networks, 2005. DySPAN 2005 (pp. 328–337).
Chapter 4
Case Study: Spectrum Management in CRN
Framework
Channel selection in the QoS provisioning framework (as discussed in Sect. 3.2),
is done through channel selection and management unit (CSMU) along with the
help of QoS control unit (QCU), policy domain and database as shown highlighted
in Fig. 4.1. The CSMU module receives the requirement of QoS from QCU, and
through the assistance of database and policy domain it selects channel. The selected
channel identifiers are then passed to the MAC protocol unit (MPU).
4.1 Spectrum Usage Behavior
As discussed in Sect. 1.6, channel selection technique depends upon various spec-
trum characteristics including channel availability, SINR, bandwidth, path loss and
primary user behavior. The primary user behavior in a cognitive operation based open
spectrum system can be classified as: deterministic and stochastic usage behavior.
4.1.1 Deterministic Usage Behavior
There are some frequency channels which inherently follow a specific pattern of
their usage over time because of the nature of the operating application [1]. This is
verified by the spectrum utilization studies done by various researchers across the
world, that have shown that the spectrum usage by incumbents follow a deterministic
time pattern over space as discussed in Sect. 1.3 and work done in [2–18]. Spectrum
occupancy measurement of Singapore is done in [19]. From this measurement and
analysis, a table is drawn and shown in Table 4.1 for some channels whose observed
spectral opportunity shows a deterministic time pattern.

112 4 Case Study: Spectrum Management in CRN Framework

stack
Application
Application QoS
Engine
Routing Module
Energy Management
QoS Control Unit Unit Packet Admission
Control
Self
Coexistence
Engine

Database domain
MAC instance
MAC Protocol Unit
MAC instance
Layer 2
and above
Layer 1
Success/Failure
report
Sensing report
Fig. 4.1 Functional modules responsible for channel selection in the QoS provisioning framework
for cognitive radio network
From Table 4.1 it can be observed that there are certain frequency bands which
are always free during certain duration of time. Local TV channels like Suria (ch12)
and Central (ch24) are closed from 12 am to 9 am and 12 am to 6 am respectively.
Therefore these channels are well suited for cognitive radio communication during
the mentioned time periods.
4.1.2 Stochastic Usage Behavior
If the spectrum usage pattern by a primary user does not follow a deterministic pattern,
it is modelled with the help of statistical learning techniques to find the hidden pattern
and to obtain the usage distribution. Several spectrum decision techniques have been
proposed in various literatures (discussed previously) and are discussed in Sect. 1.6.
4.1 Spectrum Usage Behavior 113
Table 4.1 Available frequency band over time in Singapore

Frequency (MHz) Type Free time (H)
80–88 PMR 18–12
183–186 Ch6 12–10
207–209 Ch9 18–16
212–214 Ch10 18–16
218–222 Ch11 11–06
225–230 Ch12 00–09
490–500 Ch24 00–06
614–620 Ch39 00–15
720–740 Channel 56–58 08–00
740–780 Channel 59–64 All Time
1880–1910 DECT and TDD 02–17
2270–2300 Fixed/Mobile 19–07
These works are used to model the spectrum characteristics with the assumption of
stochastic usage behavior of incumbent usage pattern on the channel. Generally these
schemes utilize the local and statistical observation to make a spectrum decision.
4.2 Reconfigurable Channel Selection
Recently, Federal Communication Corporation (FCC) has mandated the use of data-
base along with spectrum sensing to observe the spectral opportunity at a given place
over a given period of time [20]. The database assisted spectrum sensing provide an
accurate measurement of spectrum characteristics. Since the spatial and temporal
characteristics of spectral opportunities vary rapidly, an up-to-date incumbent oper-
ation database is needed for protected broadcast operation in a given location. Since
providing an up-to-date information about primary service operation is exhausting
in nature, another paradigm for information collection need to be determined. The
channels can be classified into two types depending upon their usage: Determinis-
tic time pattern channels and Stochastic channels. Therefore the system comprises
of two step reconfigurable channel selection scheme. During the first step, a chan-
nel with deterministic time pattern is selected for usage. If no such channel exists,
then stochastic channels are selected based on the requirements. The selection of
stochastic channels can be assisted with the help of database.
Figure 4.2 shows the flowchart followed by the CSMU module of the discussed
QoS provisioning framework along with the three main classes of channel selection:
Deterministic channel selection, Database assisted selection on the go and Selec-
tion on the go. Initially channel selection starts with deterministic channel selection
scheme, but if the observed spectral opportunities do not verify with that of database,
Channel Selection
Spectrum learning
Selection on the Go
Database assisted Selection

on the Go
Database
Deterministic Channel Selection
Fast spectrum
sensing
Select appropriate policy

for selected channel
Local policies
Fig. 4.2 Channel selection flow
then database assisted selection on the go method is used. In this method, statistically
learning of spectral opportunities for stochastic channel is used along with database.
If utilization of database in this methodology results in high error, then statistical
learning is used without usage of databases which is called as independent selection
on the go approach. Once a channel is selected, it is fast scanned to check out any
incumbent activity, and if it is vacant then appropriate policies are selected for the
channel. If the channel is not vacant, then deterministic spectrum learning module
updates the database. An error parameter is utilized to realize the reconfiguration of
channel selection schemes.
4.3 Channel Selection in Deterministic Environment
As discussed earlier, some channels have inherent characteristic of deterministic time

pattern of utilization. These channels are allocated ahead of time for secondary open
spectrum system so that better spectrum decisions can be made. A deterministic
channel selection scheme for channel allocation ahead of time is discussed in [21].
Some of the spectrum decision criteria discussed for QoS provisioning are: minimal
channel switching and maximal throughput of the system [22]. For the purpose
of same, three different channel selection techniques to support minimal channel
switching, maximal throughput and a value in between are discussed.
Channel selection techniques for deterministic environment employ two phases
for its operation. The first phase involves learning of the channel spectral opportu-
nity for 24 h and then based on the data obtained, the system may select a channel
4.3 Channel Selection in Deterministic Environment 115
satisfying the given constraints. These constraints could be minimal channel switch-
ing for reducing the packet delay and losses, high throughput or a value in between.
Users can also impose their own constraints to obtain the set of desired channels. The
second phase involves usage of these channels and recording the errors in channel
spectral opportunities. The errors are then used to check whether the system should
continue with the current approach or change to the selection on the go approach.
4.3.1 System Model
The system comprises of N cognitive end-users coordinated with the help of a central
controller Cognitive Access Point (CAP). Each user is equipped with a single half-
duplex transceiver, capable of performing channel sensing. The system utilizes C
non-overlapping heterogeneous primary bands each having bandwidth of Bc where
c ∈ C and C = {1, . . . , C}. These C channels are selected from a set of TV and FM
bands and are modeled as two state Markov chain as discussed in Sect. 3.4.1. A com-
mon control channel is utilized by CAP to coordinate with its users in the network;
for which any ISM band or a channel with high a availability can be selected. An
efficient spectrum sensing scheme from the literature is used at the physical layer for
spectrum sensing; and it is assumed that the result obtained by it are correct. There-
fore each CR user keeps track of the list of the available channels. This information
is passed to the CAP on common control channel.
Each CR user node uses linked list as a database for capturing these spectral
opportunities. There is a linked list assigned for each channel, therefore there are in
total C linked lists maintained by each node. The nodes of each linked list contain
the information of spectral opportunity represented by two fields: the start time and
the end time (in time-units) during which the channel is free. A typical linked list is
shown in Fig. 4.3.
The fields have the following meanings:
St : Start time from which the channel seems to be unoccupied by the primary user.
E t : End time at which the channel ceases to be unoccupied by the primary user.
The subscript t represents the node ID within a linked list and is given by t ∈
{1, . . . , T } and T is the last node ID in the linked list. The value of T may differ for
each linked list.
Ch ID
S1 E1 S2 E2 S3 E3 ST ET
NULL
Fig. 4.3 Linked list

4.3.2 Deterministic Learning with Spectrum

Selection and Usage
The system samples all C channels for 24 h with sampling interval of x minutes and
sampling time of z seconds (where z << x ∗ 60) i.e. the channel will be sensed
every x minutes for z seconds. Thus the total time required for sampling C channel
is z ∗ C/60 min (where z ∗ C/60 < x). The value of x is chosen such that the system
would not sample channels too frequently nor miss the important transitions on the
channels. The values of St and E t in Fig. 4.3 represent the time-units mapped to the
real world time with (St , E t ) ∈ {0, . . . , 1439}/sampling interval; whereas 1 time-
unit duration = sampling interval (x minutes). The terms time-unit and time-slot are
used interchangeably. If x is small, then the number of nodes in the linked list may
increase, and in the worst case scenario, the maximum number of nodes in the linked
list equals ((24 ∗ 60)/x)/2 when the channel alternates between free and occupied
states in consecutive samplings. To shorten the length of a linked list, a threshold
parameter is utilized. If the length of the spectral opportunity is greater than
(i.e. ∀t{E t − St ≥ }), then this spectral opportunity is inserted in the linked list.
For example if = 20, then if one channel is sensed free for 19 consecutive time-
units while other channel is sensed free for 22 consecutive time-slots, the availability
information of the former channel will not be entered in the linked list while that of
the latter channel will be entered.
At startup, all linked lists are empty and point to address NULL. With each sam-
pling interval and satisfaction of the threshold parameters, the nodes are updated
or inserted in the linked list. The time-unit starts from 0 (00:00 h) and increases
by 1 with each sampling interval or every x minutes. An example of the linked
list of Channel 2 after 24 h using x=10 min is shown in Fig. 4.4. It can be observed
that the Channel 2 is unoccupied during the time periods 00:00–11:40 h,
16:40–20:10 h and 23:00–24:00 (00:00); and is occupied during the time periods
11:40–16:40 h and 20:10–23:00 h.
The sampling continues for 24 h, at the end of which the system would have obtained
spectral opportunities for all channels. This information is passed to CAP, which
consolidates the information received from all the nodes to obtain the global system
spectral opportunity. This consolidation is required so that every node in the system
agrees on the availability of the channel. A simple method to obtain the global channel
opportunity is to convert all the linked lists into matrices and compute the intersection
on them, thus reducing the computation complexity.
The linked lists received by CAP from each node are described as L L i, j where
i ∈ C , j ∈ N and N = {1, . . . , N }. Let η = (24 ∗ 60/x) − 1 and η = {0, . . . , η}.
Let the matrix for the linked list received from node j be MAT( j), where j ∈ N . The
Ch ID = 2
0 69 100 120 138 143 NULL
Fig. 4.4 Channel usage pattern of Channel 2 for one day (24 h)
row and column of the matrix represent channels and time-units respectively. The
conversion from linked lists to this matrix can be done as follows. In the following
conversion index j is dropped.

1, ∃t[k ∈ {St . . . E t }]
MATi,k = ∀i ∈ C , ∀k ∈ η , t ∈ T (4.1)
0, otherwise
The predicate logic ∃t[k ∈ {St , . . . , E t }] or ∃t[St ≤ k ≤ E t ], evaluates to true

whenever the time-unit value falls within the node field St and E t values. Therefore
the elements of the matrix are either 0 or 1 representing a PU is present or not
respectively on a channel corresponding to the given time-slot for the node j. Overall,
N such matrices are obtained. To consolidate all the global information, the CAP
performs AND operation on all the corresponding locations of the matrices to obtain
a channel availability matrix satisfying the channel availability of all the nodes in
the system, i.e.

j=N
Mi,k = MATi,k ( j), ∀i ∈ C and ∀k ∈ η (4.2)
j=1
The channel availability matrix M agrees with all the CR nodes present in the system
with respect to the availability of the channel. An example matrix is shown in Fig. 4.5.
It is assumed that at any given time-slot (time-unit) at least one channel is available
for communication. Hence
∀k∃i{Mi,k = 1}, where k ∈ η , i ∈ C (4.3)
Different users have different requirements for the channel selection. The matrix M
can be processed to obtain the desired channel or set of channels satisfying the users
constraints. If a single channel is required at a given time-slot, then the solution is
represented with a vector. The vector index is represented by the time-slot while its
content represents the channel ID. Therefore this vector guides the system to choose
a given channel at a given time-slot. A typical vector could be initialized as:
Fig. 4.5 A typical example Time Units (k)

0 1 2 3 4 5 6 ... η
of matrix showing
availability of channels in 1 0 0 0 0 1 1 0 0
CR network
2 1 1 1 1 0 0 0 1
Channel ID (i)
...
3 0 1 1 1 0 0 0 1
4 0 0 0 0 1 1 1 0
...
...
...
C 0 1 1 1 1 0 0 ... 1
Vk = {i : ∀k∃i[Mi,k = 1]}, where ∀i ∈ C and ∀k ∈ η (4.4)
Multiple vectors may satisfy this equation and hence multiple solutions may exist
if no constraints are applied. The above vector simply chooses a channel which
is available during a given time-slot and thus represents the solution space for all
possible vectors. A user may apply single or multiple constraints on the system to
impose restriction on channel selection. That is, the solution vector would be an
element from the solution space set obtained in Eq. 4.4. Three different strategies
of channel selection are discussed so as to meet the constraints of: minimal channel
switching, maximal throughput and a value in between.
4.3.3 Minimal Channel Switch Requirement
Channel switching is costly for a radio device especially in CR devices [23]. Real-
time applications suffer from packet losses, jitters and delays induced by frequent
channel switching in CRN. Therefore a comparison of different channel switching
techniques indirectly corresponds to a comparison of different packet delays, losses
and jitters induced by channel switching in the network. Minimal channel switching
refers to the selection of channels such that the system would stay on that channel
for a long period of time without hopping onto other channels. The problem can be
formulated as to obtain a solution vector VMIN such that when the system follows the
solution vector, the overall switching is minimized for the entire duration. Therefore
the problem is to find VMIN which has the least total number of channel switching
as compared to other vectors available, i.e.,
η−1 η−1

Switching(VMINk ) ≤ Switching(Vk ) (4.5)
k=0 k=0
where ∀V ∈ {set of all possible vectors V obtained from Eq. 4.4} and Switching()
is a binary function defined as:

1, if X i = X i+1
Switching(X i ) = (4.6)
0, otherwise
where X is channel vector and i is its time index.

A solution to this problem can be obtained by using a greedy approach. This
approach selects the channel which has the maximum period of continuous vacancy
over a time interval. Algorithm 4.1 gives the implementation detail of the greedy
approach. The algorithm first initializes the vector to value zer o (line 1). If the
solution vector contains zer o at any time-slot, it means there are no channels available
during that time-slot. The outer loop is used to index the time-slots (lines 3–18), the
middle loop is used to check out all the channels (lines 4–14) and the inner most loop
is used to check out the length of the continuous vacancy (lines 7–11). The variable
array trail stores the length of continuous vacancy over all channels. At the end,
another loop is used to check which channel has the longest spectral opportunity and
is stored in VMIN (lines 15–17).
Lemma 4.1 If on average there are θ channels available per time-unit, then the
total number of vectors V available is θη+1 . The average is obtained over space or
location.
Algorithm 4.1: Algorithm for minimal channel switching using greedy

approach. The max() function returns the maximum value among all elements
of a vector and argmax() function returns the argument for which the function
value is maximum.
1 VMINk ←− {0}, ∀k ∈ η ;
2 traili ←− {0}, ∀i ∈ C ;
3 for k = 0 to η − 1 do
4 for i = 1 to C do
5 traili ← 0;
6 if Mi,k = 1 then
7 for d = k + 1 to η do
8 if Mi,d = 0 then
9 break-for;
10 end
11 end
12 traili ← d − k
13 end
14 end
15 for k = k to k + max(trail) do
16 VMINk = argmax j∈C (trail);
17 end
18 end
Proof The total number of vectors V equals to the total number of combinations
through which the system can find at least one channel available in each time-slot.
Since on average there are θ channels available at each time-slot, the total number of
combinations available from time-unit = 0 to time-unit = η, equals θ.θ.θ . . . η + 1
times i.e. θη+1 . Thus, the search space is exponential.
Lemma 4.2 The algorithm based on the greedy approach of selecting the longest
consecutive free time-slots provides an optimal solution to the minimal channel
switching problem. In other words there exist no other vector U that has smaller
number of channel switching than that obtained by the Algorithm 4.1. The proof is
given in Appendix A.
4.3.4 Maximum Throughput Requirement
Some bandwidth intensive applications like peer-to-peer (P2P) file sharing require
high throughput across 24 h. Normally, these applications are unconstrained from
the perspective of channel selection and the only requirement is high bandwidth.
Therefore, an un-constrained channel selection technique with the aim of maximizing
throughput across 24 h is provided. Maximum throughput tends to provide channel
selection strategy in such a way that the system will select those channels which
provide highest bandwidth. Therefore the solution vector to this problem should
provide the list of channels which would provide maximum bandwidth over a given
time-slot. The maximum throughput vector VTH can be calculated as shown in
Algorithm 4.2.
The algorithm first sorts the IDs of the channel in such a way that the lowest index
has the highest bandwidth. The mapping is done internally from the older index to
the newly assigned index, and is not shown here. The algorithm then selects the
channel which has maximum bandwidth over a given time-slot and enter its ID into
the vector VTH.
Algorithm 4.2: Algorithm for obtaining maximum throughput vector

1 Rearrange the rows representing the channels in the matrix M such that lowest index has the
highest bandwidth i.e. Bi > Bi+1 , ∀i ∈ C ;
2 VTHk ←− {0}, ∀k ∈ η ;
3 for k = 0 to η do
4 for i = 1 to C do
5 if Mi,k = 1 then
6 VTHk ← i;
7 break-for;
8 end
9 end
10 end
4.3.5 Intermediate Solution to Provide High Throughput

Along with Minimal Channel Switch Requirement
The previous two approaches provide either the minimum number of channel switch-
ing or the highest throughput. Sometimes, however it is required to have a solution
in between i.e. to provide a vector consisting of channel IDs such that if the system
follows the list, then it would provide an intermediate solution which lies in-between
the two approaches. Throughput loss caused by channel switching is not considered,
as switching to a channel having a high bandwidth can easily compensate for the
loss. Therefore the aim of this channel selection mechanism is to find a solution
vector such that it provides bounded channel switching (for bounded packet loss, jit-
ter and delay induced by channel switching) along with the guaranteed throughput.
This problem lies in the domain of NP-complete problems, and called as the High
Throughput Minimal Switching (HTMS) problem.
Complexity Analysis of HTMS Problem
The optimization problem of HTMS is converted into decision problem to define its
complexity. Every optimization problem can be converted to a decision problem [24],
and the decision problem corresponding to HTMS problem is defined as follows:
Definition 4.1 Given a channel availability matrix M, and numbers ω and δ, does
there exist a vector V which satisfies all the constraints of the HTMS problem along
with the condition that total number of switching <δ and bandwidth >ω?
Lemma 4.3 HTMS decision problem is in NP. The proof is given in Appendix B.
Lemma 4.4 HTMS decision problem is NP-Hard. The proof is given in Appendix C.
Theorem 4.1 HTMS decision problem is NP-complete problem.
Proof From Lemmas 4.3 and 4.4, it can be concluded that HTMS problem is NP-
complete problem.
Heuristic Algorithm for HTMS problem

Since HTMS problem is NP-complete problem, a simple heuristic approach for its
sub-optimal solution is provided in Algorithm 4.3. The algorithm first selects the
minimal channel switching vector and then tries to satisfy the goal of guaranteed
bandwidth. This is done by selecting the channels which were previously not selected
and have higher bandwidth than that of the currently selected channel. The allowed
switching and guaranteed bandwidth is represented by δ and ω respectively.
The algorithm first obtains the minimal channel switching vector VMIN from
Algorithm 4.1. It then updates the vector with the channel IDs which have higher
bandwidth than the currently selected channels in VMIN (lines 5–12). The while
loop iterates until all channels have been checked (lines 14–16) or the bandwidth
desired is obtained (line 3) and hence guarantees the convergence.
4.3.6 Complexity Analysis of Algorithms
The worst case scenarios for all algorithms are observed. Algorithm 4.1 has one
parent f or loop (lines 3–18) with two child f or loops (lines 4–14 and lines 15–17).
The running length of the parent loop is η and that of child loops are C and α, whereas
the parameter α is variable and its value depends on the parameter η. The child loop
(lines 4–14) with running length C is nested with another child loop (lines 7–11) with
Algorithm 4.3: Algorithm for intermediate channel selection for HTMS prob-
lem. The function band_sum() returns the overall bandwidth of the system.
1 Obtain VMIN from Algorithm 4.1

2 i = 1;
3 while ω ≤ band_sum(VMIN) do
4 f lag = 0;
5 for k = 0 to η do
6 if Mi,k = 1 & VMINk > i then
7 VMINk = i ;
8 end
9 if ω ≤ band_sum(VMIN) then
10 flag=1; break-for;
11 end
12 end
13 i = i + 1;
14 if i = C || f lag = 1 then
15 break-while;
16 end
17 end
running length α. The outermost parent loop (line 3) is complemented with the help
of innermost for loop (line 7) and the other child for loop (line 15). Thus the running
length of outermost loop is complemented. The complexity of max() function and
argmax() function for current application is assumed to be O(n), where n is the
length of the vectors for these functions. Thus the complexity obtained of Algorithm
4.1 is O(η(C + n)) or O(η 2 ).
Algorithm 4.2 has two nested f or loops with running length of η and C. Thus
the complexity of the algorithm is O(ηC) or O(η 2 ).
Algorithm 4.3 obtains the vector VMIN using Algorithm 4.1 and then updates the
vector to satisfy the bandwidth guarantee. The outer while loop has running length
of C and the inner f or loop has running length of η + 1. Thus the complexity of this
algorithm equals, the complexity of Algorithm 4.1 +O(C(η + 1)) or O(η 2 ).
4.3.7 Spectrum Usage
When the system has obtained the spectral opportunities for 24 h and channel selec-
tion vectors (following the user constraints), the selected channel can be utilized
for communication purposes. Spectrum usage consists of normal communication
between nodes in CRN using the channels provided by the vectors obtained with dif-
ferent channel selection algorithms. An error counter ξi , ∀i ∈ C is associated with
each channel. Each time before a packet is transmitted, fast sensing [25] is performed
on the channel to detect the presence of any PU. If a PU is present on the channel, then
that information is updated in the global channel availability matrix M and the error
counter associated with that channel is increased by 1. If the error counter reaches
the threshold , then that channel (row) information is removed from the channel
availability matrix M. If the total error on all channels crosses the threshold χ, then
the system switches onto Selection on the go approach. The parameters and χ are
user selectable and depends upon the channel environment.
4.3.8 Performance Analysis
Since deterministic time pattern of the channels are being utilized, this model is
initialized for the first 24 h of the system, after that the recorded state is used for the
next 24 h. The given scheme is evaluated using discrete event simulation in Matlab
which simulates the network up to frame level. The channel availability parameters
(discussed in Sect. 3.4.1) α and β are set to 0.1, so as to model the channels whose
availability do not change abruptly. A low value of α and β means that the system
is likely to stay either in ON state or in OFF state most of the time without state
transition. The bandwidth (data rate) of the channel is modelled as to be linear with
the number of channels used, i.e. the minimum bandwidth of a channel is given a
value of 2 Megabit per minute (Mbpm) and the maximum bandwidth of the channel
is 2 ∗ C, where C is the maximum number of channels used for the simulation.
Therefore the bandwidth is distributed linearly from Channel 1 to Channel C as 2 to
2 ∗ C Mbpm. The error counter ξi , ∀i ∈ C is set to 10.
The discussed techniques are compared against three other channel selection tech-
niques: stochastic channel selection (SCS) [26], automatic channel selection (ACS)
[27], and minimum variance-based spectrum decision single selection (MVSD-SS)
[23]. SCS [26] utilizes learning automata and aims at reducing the number of chan-
nel switching by selecting a channel which maximizes the probability of successful
transmission. ACS [27] aims at selecting a channel that provides guaranteed band-
width and fairness to the system with the help of measured primary user activity on
the channel. MVSD-SS [23] aims at selecting a channel that increases the bandwidth
utilization of the system. For a fair comparison, some of the constraints have been
removed which were utilized in the literature. The given channel selection techniques
utilize the infrastructure version of the media access scheme discussed in Chap. 5.
Each data point in the graph is an average of 100 runs while the unit of bandwidth
and sampling interval are Mbpm and minute respectively.
The algorithms is evaluated on the basis of following performances:
1. Overall Bandwidth of the system
2. Overall Channel switching required.
Overall bandwidth of the system refers to the bandwidth obtained by the system in
24 h. At some point the bandwidth obtained might be less and sometime might be
the high. The bandwidths are consolidated by summing them up for 24 h. Overall
channel switching required refers to the total switching needed for the 24 h operation
of the system.
300
No. of channel switching required

250 Algorithm 4.2
Stochastic Channel Selection
200 Automatic Channel Selection
MVSD−SS
150 Algorithm 4.1
100
50
0
1 2 3 4 5 6 7 8 9 10
Sampling Interval
Fig. 4.6 Number of channel switching required versus sampling interval (x minutes)
Figure 4.6 shows sampling interval versus number of channel switching required
for overall duration of 24 h with C = 10. It can be seen from the graph that Algo-
rithm 4.1 provides minimal channel switching which is better than SCS technique.
Although SCS technique aims at minimizing the channel switching, it lacks the
knowledge of spectral opportunity ahead of time and thus suffers from degraded per-
formance. All other techniques do not consider channel switching in their decision
making process and therefore suffers a large amount of channel switching. Figure 4.6
also implies that Algorithm 4.2 is not good for minimal channel switching as it has
the worst performance.
Figure 4.7 shows the bandwidth obtained versus sampling interval with C = 10.
Results show that the bandwidth of the system is independent of the sampling interval
used. It is also noted that Algorithm 4.2 provides the highest bandwidth among all
4
x 10
2.8
2.6
Bandwidth obtained
2.4
Algorithm 4.2
2.2 Stochastic Channel Selection
Automatic Channel Selection
2
MVSD−SS
1.8 Algorithm 4.1
1.6
1.4
1 2 3 4 5 6 7 8 9 10
Sampling Interval
Fig. 4.7 Bandwidth versus sampling interval (x minutes)

30
25
No. of channel switching required Algorithm 4.2

20 Stochastic Channel Selection
MVSD−SS
Algorithm 4.1
15
10
0
5 10 15 20 25 30
No. of channels
Fig. 4.8 Number of channel switching required versus number of channels utilized in the system
other techniques. This is the maximum unconstrained bandwidth that can be obtained
by the system. Despite the removal of constraints from ACS and MVSD-SS to select
high bandwidth channels for a fair comparison, these techniques do not have the
knowledge of spectral opportunities ahead of time and suffer from performance loss.
SCS aims at reducing the number of channel switching and does not account for
throughput. Therefore its performance is near to that of Algorithm 4.1.
Figure 4.8 shows the number of channel switching required versus number of
channels in the system with sampling interval = 10 min. It might be noted that
the number of channel switching required in Algorithm 4.2 and all other techniques
except Algorithm 4.1 is insensitive to the number of channels used in the system. The
number of channel switching decreases with an increase in the number of channels in
the system for Algorithm 4.1. With an increase in number of channels in the system,
the number of combinations of vector V increases in the channel availability matrix
and hence the switching decreases.
Figure 4.9 shows bandwidth obtained versus number of channels used in the sys-
tem with sampling interval = 10. Note that Algorithm 4.2 obtains the highest band-
width while other techniques (ACS and MVSD-SS) perform marginally poorer. The
difference becomes significant when the number of channels used in the system
becomes large. The performance of the Algorithm 4.1 and SCS technique is simi-
lar. The bandwidth of each scheme increases linearly because the bandwidth of each
channel also increases linearly with the number of channels in the system (bandwidth
is distributed linearly from channel 1 to channel C as 2 to 2 ∗ C Mbpm).
Figure 4.10a, b show the performance for Algorithm 4.3 with guaranteed through-
put (bandwidth) and maximum allowable switching. For the purpose of generaliza-
tion, the rule of thumb “more than half” is used and assumed that the number of
4
x 10
8
Algorithm 4.2
7 Stochastic Channel Selection
6 MVSD−SS
Algorithm 4.1
5
Bandwidth
0
5 10 15 20 25 30
No. of channels
Fig. 4.9 Bandwidth versus number of channels
switching allowed per slot = 0.1 and total bandwidth available per slot = 60 % of
the maximum bandwidth of the channel. The parameters α and β are set to 0.1.
Figure 4.10a shows number of channel switching allowed versus number of chan-
nels in the system with sampling interval = 10. The figure shows that the algorithm
obtains channel switching lower than the allowed value. In addition the graph shows
that there is no particular relation between the number of channels used in the system
with the number of channel switching required.
Figure 4.10b shows bandwidth obtained in the system versus number of channels
used in the system with sampling interval = 10. Note that Algorithm 4.3 satisfies
the bandwidth guarantee required by the system. The bandwidth of each scheme
increases linearly because the bandwidth of each channel also increases linearly
with the number of channels in the system.
Figure 4.10c shows number of channel switching allowed versus sampling inter-
val for an overall duration of 24 h with C = 10. Algorithm 4.3 provides chan-
nel switching lower than that of the allowed value. The allowed switching (δ)
decreases with the increase in sampling interval value and so is the value obtained by
Algorithm 4.3.
Figure 4.10d shows bandwidth obtained versus sampling interval with C = 10.
Note that the algorithm guarantees the minimum bandwidth to the system. Here
the bandwidth obtained from Algorithm 4.3 increases in an up down fashion. The
randomness is due to the random channel availability and the increase in overall band-
width is due to the selection of channels having high bandwidth and high switching.
Figure 4.10c shows a decrease in the difference between the switching obtained by
Algorithm 4.3 and the allowed switching, and is accounted for increase in bandwidth
as shown in Fig. 4.10d.
No. of switching
15
10
5
5 10 15 20 25 30 35 40 45
(a) No. of channels
4
x 10
Bandwidth (Mbpm)
10
0
5 10 15 20 25 30 35 40 45
(b) No. of channels
No. of switching
200
100
0
1 2 3 4 5 6 7 8 9 10
(c) Sampling interval
4
x 10
Bandwidth (Mbpm)
1.9
1.8
1.7
1 2 3 4 5 6 7 8 9 10
(d) Sampling interval
Algorithm 4.3 Allowed switching Desired bandwidth
Fig. 4.10 Performance analysis of Algorithm 4.3 with respect to desired bandwidth and allowed
switching
4.3.9 Ad-hoc Mode Operation
The channel selection techniques discussed in the previous section assumed the sys-
tem model of Single Network in Infrastructure Mode. The discussed channel selection
techniques can be modified to apply for networks operating in ad-hoc mode. To do
so, additional protocol is required to achieve the goal of global channel availabil-
ity matrix which is same for all the CR users in the ad-hoc network. If every CR
user has same channel availability matrix, then the channel selection algorithm with
desired constraints operating over all CR users will result in common channel selec-
tion, which can be utilized by an ad-hoc media access scheme for communication
purposes.
4.4 Channel Selection in Stochastic Environment
If the error counter associated with the channel crosses a threshold (ξ), then that
channel is classified as stochastic channel. If the total error counter for all channels
crosses a given threshold (), then the system switches onto Selection on the go
approach. The system may utilize any of the stochastic channel selection technique
available in the literature. Most of these channel selection technique are based on
statistical learning, which introduces complexity to the system. Here a stochastic
channel selection technique is discussed which map packets to channels on the basis
of a tabular mapping [28]. This tabular mapping reduces the complexity of the system
as well as facilitates fast channel selection.
A generalization is made regarding the user capability of multi-tasking and run-
ning multiple different applications, and hence it can be assumed that a user can
generate traffic with multiple priorities. Therefore a user is not classified as real time
or non-real time, instead data packets are classified as real time or non-real time.
Here, a spectrum decision scheme for cognitive radio network in stochastic channel
environment is discussed which divides the data traffic into four priorities on the
basis of the QoS requirements and channels into eight categories on the basis of their
bandwidth, channel availability and primary user (PU) fluctuation (discussed later).
4.4.1 System Model
The system comprises of N open spectrum CR users and utilizes a common control
channel which may be selected from one of the ISM bands or a channel with high
availability. Each of the C channels utilized in the system has a bandwidth of Bi ,
∀i ∈ {1, . . . , C}. Each channel is associated with an ID and counter for its identifi-
cation and load balancing purposes (discussed later) respectively. Therefore a total
of C counters is maintained by each node in the network. Each CR user is in com-
munication range with all other CR users and the transmission range of a PU is more
than that of the CR user.
As discussed in Chap. 3, the user traffic is classified into four different priorities
as shown in Table 3.1. It is assumed that the application layer is capable of tagging
(marking) the packets associated with a particular traffic. The channel availability
is modelled as two-state Markov chain as discussed in Sect. 3.4.1. It is assumed
that each CR user statistically learns the channel transition probabilities from their
previous sensing. Therefore each CR user knows the channel transition probabilities
for all C channels.
4.4 Channel Selection in Stochastic Environment 129
4.4.2 Communication Segment
A communication segment (t, r ) is defined as a pair of transmitting and receiving

CR users for a session. A communication segment operates over a single channel
and does not spawn over multiple channels during a session. Also a communication
segment can carry only a single class (type/priority) of data and therefore has the
priority depending upon the type of data traffic it is carrying. This section will use
the word ‘connection’ interchangeably with the word ‘communication-segment’.
4.4.3 Spectrum Decision and Mapping of Packets
The spectrum decision involves selection of channels by CR users depending upon

the data packet needed to be sent. All C channels are classified into eight types on
the basis of bandwidth, channel availability and PU fluctuation. The calculation of
channel availability and fluctuation is discussed later. Table 4.2 shows the tagging of
these channels. The term ‘High’ and ‘Low’ are subjective and may vary depending
upon application and licensed channels. Voice packets need those channels which
have higher availability and low fluctuation. An example would be if α = 0.9 and
β = 0.9; the channel is although available with probability of 0.5 but it alternates at
each time slot, resulting in jittering at each step. Therefore PU fluctuation need to be
taken care of by using fluctuation parameter. The fluctuation parameter δ is defined
as:
αi + βi
δi = ∀i ∈ C (4.7)
2
The fluctuation parameter δ ∈ [0, 1] tells how much the PUs’ presence fluctuate
(varies) on the given channel. A higher value of δ represents high fluctuation while
Table 4.2 Tagging and classification of primary licensed channels

Bandwidth Channel availability Primary user fluctuation Tag
High High High 0
High High Low 1
High Low High 2
High Low Low 3
Low High High 4
Low High Low 5
Low Low High 6
Low Low Low 7
Table 4.3 Mapping of data packets to channel requirements

Access category Bandwidth Channel availability Primary user
fluctuation
Voice High/Low High Low
Video High High High/Low
Best effort Low High/Low Low
High High/Low High/low
Background Low High/Low High/Low
a low value of δ represents low fluctuation. Fluctuation parameter only represent the
PU fluctuation and does not represents the actual availability of the channel.
Therefore high availability of the channel and minimal fluctuation of the PU on
that channel is desired for the system supporting real time traffic like voice. A quick
analysis of Eqs. 3.2 and 4.7 indicates that these criteria can be met with low value of
β and equal value of α and β. The low value of β decreases the fluctuation of the PU
on the channel while an equal value of α and β increases the channel availability.
Once the quantity ‘High’ and ‘Low’ are associated with bandwidth, availability of
all the channels along with the PUs fluctuation on it, the CSMU can define mapping
function for mapping packets to channels. The mapping function is defined in Table
4.3. The table shows that the voice packets can choose channels with low bandwidth,
as the bandwidth of voice packet is small. The guarantee is made in terms of channel
availability and PU fluctuation on the channel. Similarly, video packets can choose
a channel with high PU fluctuation as videos can be buffered if needed. Background
traffic can only choose channels with low bandwidth, whereas Best Effort data con-
siderably overlaps with other data priorities. The reason is that if video or voice
traffic is not present then best-effort traffic can use the channel. Each traffic starts
with selecting low bandwidth channel in its class i.e. a voice and best-effort traffic
will first try to choose a low bandwidth channel and if not available then move to
next higher bandwidth channel.
4.4.4 Load Balancing
Since there are eight categories of channel as shown in Table 4.2, the numbers of
channel in a single category is represented by γi , ∀i ∈ C . Therefore i=0
7
γi = C.
It may be possible that some category may not have any channels in them. Let ξk ,
∀k ∈ {0..7} represent the set of channels IDs in a given category. The purpose of the
load balancing system is to uniformly distribute the communication segments within
the channels of a given category. Algorithm 4.4 provides the method of load balancing
to the system while doing the spectrum decision for a communication segment. The
index for ξ is dropped because the algorithm represents the load balancing within a
category. The variable array cntr maintains the list of counter for all the channels,
and it is used to count the activities on a channel. The counter value for each channel
is set to zero during the initialization of the system. The sensing report unit of the
framework updates the value of the counter based on the external activity (PU or
any other CR user) whereas the counter variable is also updated inside the algorithm
block to count for the internal activity. The higher the value of the counter for a
channel, the higher the activities it has got on it. The goal of the algorithm is to select
a channel within a given category of channels such that it has minimal activity on it. If
a channel with minimal number of activity is selected and used, a report is generated
with the help of success/failure unit of the framework (lines 6–8). If the success
reported is negative, then that channel is removed from the category for the current
session and minimal activity channel is selected from the rest of the channels (lines
9–12). If all channels are removed then the algorithm starts from the beginning. The
counter is reset to 0 once it reaches a threshold. The threshold can vary according to
the application.
The terms ‘High’ and ‘Low’ for bandwidth of a channel, channel availability and
PU fluctuation on a given channel are defined in order to obtain the performance
analysis. Any channel having bandwidth greater than or equal to average bandwidth
is called as ‘High’ bandwidth or otherwise ‘Low’ bandwidth. The average bandwidth
is calculated using Eq. 4.8.
C
Bi
avg_band = i=1 (4.8)
C
Algorithm 4.4: Algorithm for load balancing in a given category of channels

1 Initialize :ψ = ξ;
2 Label :
3 if ψ = {} then
4 ψ = ξ;
5 end
6 ch_id = argmin i∈ψ (cntri );
7 cntri = cntri + 1;
8 r epor t = success_r epor t (ch_id);
9 if r epor t = 0 then
10 ψ = ψ − {ch_id};
11 goto Label;
12 end
Therefore,

H igh, If Bandwidth i ≥ avg_band
Bandwidth i = (4.9)
Low, otherwise
Channel availability and PU fluctuation is similarly assigned as ‘High’ and ‘Low’ as

follows. C
i
avg_ch_av = i=1 (4.10)
C

H igh, If i ≥ avg_ch_av
Channel Availabilit yi = (4.11)
Low, otherwise
C
i=1 δi
avg_ pu_ f l = (4.12)
C

H igh, If δi ≥ avg_ pu_ f l
Pu Fluctuation i = (4.13)
Low, otherwise
The discussed channel selection scheme is implemented on top of the MAC protocol
discussed in Chap. 5. The discussed scheme is evaluated using discrete event sim-
ulation in Matlab which simulates the network up to frame level. The performance
of the discussed framework is evaluated in terms of connection fail ratio, number of
connections failed and number of channels used. Connection fail ratio is calculated
by
number of failed connections
Connection fail ratio =
number of connections initiated
The parameters α and β are assumed to be uniformly distributed over the range [0,
1] i.e. the transition probabilities of the channels are uniformly distributed over the
range [0, 1]. The bandwidth of the channel is modelled to be linear with the number
of channels used, i.e. the minimum bandwidth of a channel is given a value of
50 Kbps and the maximum bandwidth of the channel is 50 ∗ C Kbps, where C is the
maximum number of channels used for the simulation. Therefore the bandwidth is
distributed linearly from channel 1 to channel C as 50 to 50 ∗ C Kbps.
Figure 4.11 shows the graph of connection fail ratio versus the total number of
channels used the system. The number of voice communication segment, video
communication segment, best-effort communication segment and background traffic
communication segment used are 2, 5, 30 and 13 respectively giving the percentage
ratio of 4, 10, 60 and 26 % for voice, video, best-effort and background traffic com-
munication segments respectively. The number of channels is varied from 10 to 40.
The connection fail ratio decreases with increase in number of channels for all types
of communication segments. Since the number of voice traffic is small and the num-
ber of channels is adequate to satisfy the requirement, the connection fail ratio for
0.9
Voice data
0.8
Video data
0.7 Best effort data
Connection fail ratio Background traffic
0.6
0.5
0.4
0.3
0.2
0.1
0
10 15 20 25 30 35 40
No. of channels
Fig. 4.11 Connection fail ratio versus number of channels
voice traffic is smallest. Background traffic shows higher blockage at start because
most of the channels are used by the best-effort traffic (since there is overlapping of
the channels for these two traffic types in mapping Table 4.3).
Figure 4.12 shows the graph of number of failed connections versus total num-
ber of communication segments for C = 50. The total number of communication
segments comprises 4, 10, 60 and 26 % of voice, video, best-effort and background
traffic communication segments respectively. From the graph it is clear that the num-
120
Voice data
100
Video data
Best effort data
No. of channels used
80 Background traffic
60
40
20
0
20 40 60 80 100 120 140 160 180 200
Total no. of communication segments
Fig. 4.12 Number of connections failed versus total number of communication segments
30
25
Voice data
No. of channels used
20
Video data
Best effort data
15 Background traffic
10
0
20 40 60 80 100 120 140 160 180 200
Total no. of communication segments
Fig. 4.13 Number of channels used versus total number of communication segments
ber of connections will fail as the number of connections (communication segments)

increases. A steep increase can be seen when total number of communication seg-
ments equals 60, which can be explained by the fact that channels become busy
after this point leading to failure in new connections. There is steep increase in
the best-effort segment losses because of the large share of the total segments and
some of the channels mapped to this traffic type is used by higher order traffic too
(Table 4.3).
Figure 4.13 shows the load balancing in the system by taking in to account the
number of channels used by a traffic type with respect to total number of commu-
nication segments. The total number of communication segments comprises 4, 10,
60 and 26 % of voice, video, best-effort and background traffic communication seg-
ments respectively. The total number of channels used is 50. It is clear from the graph
that the number of channels used for all traffic types increases till total number of
communication segments equals 60. After which there are no channels to fulfil the
upcoming requests. As the number of communication segment increases, the num-
ber of voice and video data communication segments also increase leading to more
use of the channels mapped by voice and video data. Thus there is decrease in the
channel use by the web traffic data which were overlapped with the voice and video
data (Table 4.3).
4.5 Discussion 135
4.5 Discussion
The performance of a network depends upon the bandwidth and quality of the wireless
channel in which it operates. A channel can be characterized on the basis of its
availability; and on the basis of observation, channels may have deterministic/non-
deterministic usage pattern. Channels like TV and FM bands can be utilized for the
purpose of modelling deterministic behavior of licensed users. If channels do not
have any deterministic usage pattern, then stochastic channel selection techniques
(Selection on the go approach) are utilized.
In this chapter we have discussed a reconfigurable channel selection technique
which initially starts with deterministic channel selection techniques for channels
with deterministic usage behavior. If channels do not follow a deterministic behav-
ior, the increase in error counter forces the system to utilize another discussed channel
selection technique for stochastic environment. Three different algorithms have been
discussed in deterministic channel environment targeting different aspects. These
algorithms solve the problem of minimal channel switching and high throughput.
Channel switching is reduced to mitigate the effects of packet losses, delays and
jitter caused by channel switching. One of the discussed algorithm provides an inter-
mediate result between minimal channel switching and maximum throughput, solv-
ing the high throughput minimal channel switching problem. In stochastic channel
environment, the given technique divides channel into into eight different categories
based on bandwidth, channel availability and PU fluctuation parameter. A mapping is
provided from the data traffic priority to the channel type for the purpose of channel
selection, along with a load balancing algorithm to balance the traffic on different
channels.
Performance analysis of channel selection techniques in deterministic environ-
ment is done with three other channel selection techniques. Simulation result shows
that the performance of the discussed algorithms based on the given system model
approach provides better performance of the overall system. In addition the given
model can provide a trade-off between minimal channel switching and maximum
bandwidth. Similarly, in stochastic environment, the performance analysis shows that
the channel selection can be utilized for obtaining different priorities with effective
load distribution. Although this chapter covered the schemes for spectrum selec-
tion in deterministic and stochastic channel environment, it misses the important
parameter of energy consumption.
References
1. Mishra, V., Tong, L. C., & Syin, C. (2013). Deterministic time pattern based channel selection
in cognitive radio network. In The International Conference on Information Networking 2013
(ICOIN) (pp. 169–174).
2. Erpek, T., Lofquist, M., & Patton, K. (2007). Spectrum occupancy measurements loring com-
merce centre limestone, Maine September 18–20, 2007. Shared Spectrum Company Report.
3. Erpek, T., Steadman, K., & Jones, D. (2007). Dublin Ireland spectrum occupancy measurements
collected on April 16–18, 2007. Shared Spectrum Company Report.
4. McHenry, M. A., McCloskey, D., Roberson, D., & MacDonald, J. T. (2005). Spectrum occu-
pancy measurements Chicago, Illinois November 16–18, 2005. Shared Spectrum Company
Report.
5. McHenry, M. A., & Steadman, K. (2005). Spectrum occupancy measurements location 1 of 6:
Riverbend park, Great Falls, Virginia April 7, 2004. Shared Spectrum Company Report.
6. McHenry, M. A., & Steadman, K. (2005). Spectrum occupancy measurements location 2 of 6:
Tyson’s square center, Vienna, Virginia April 9, 2004. Shared Spectrum Company Report.
7. McHenry, M. A., & Chunduri, S. (2005). Spectrum occupancy measurements location 3 of 6:
National science foundation building roof April 16, 2004. Shared Spectrum Company Report,
Revision 2.
8. McHenry, M. A., & McCloskey, D. (2005). Spectrum occupancy measurements location 4 of
6: Republican national convention, New York City, New York August 30, 2004–September 3,
2004. Shared Spectrum Company Report, Revision 2.
9. McHenry, M. A., & Steadman, K. (2005). Spectrum occupancy measurements location 5 of
6: National radio astronomy observatory (NRAO) Green Bank, West Virginia October 10–11,
2004. Shared Spectrum Company Report, Revision 3.
10. McHenry, M. A., & Steadman, K. (2005). Spectrum occupancy measurements: Shared spectrum
building roof, Vienna, Virginia, December 15–16, 2004. Shared Spectrum Company Report.
11. Valenta, V., Marsalek, R., Baudoin, G., Villegas, M., Suarez, M., & Robert, F. (2010). Survey
on spectrum utilization in europe: Measurements, analyses and observations. In Proceedings
of the Fifth International Conference on Cognitive Radio Oriented Wireless Networks Com-
munications (CROWNCOM) (pp. 1–5).
12. Patil, K., Skouby, K., Chandra, A., & Prasad, R. (2011). Spectrum occupancy statistics in the
context of cognitive radio. In 14th International Symposium on Wireless Personal Multimedia
Communications (WPMC) (pp. 1–5).
13. Sanders, F. (1998). Broadband spectrum surveys in Denver, CO, San Diego, CA, and Los Ange-
les, CA: Methodology, analysis, and comparative results. In IEEE International Symposium on
Electromagnetic Compatibility (Vol. 2, pp. 988–993).
14. Chiang, R., Rowe, G., & Sowerby, K. (2007). A quantitative analysis of spectral occupancy mea-
surements for cognitive radio. In IEEE 65th Vehicular Technology Conference, 2007. VTC2007-
Spring (pp. 3016–3020).
15. Wellens, M., Wu, J., & Mahonen, P. (2007). Evaluation of spectrum occupancy in indoor and
outdoor scenario in the context of cognitive radio. In 2nd International Conference on Cognitive
Radio Oriented Wireless Networks and Communications, 2007. CROWNCOM (pp. 420–427).
16. Yin, S., Chen, D., Zhang, Q., Liu, M., & Li, S. (2012). Mining spectrum usage data: A large-
scale spectrum measurement study. IEEE Transactions on Mobile Computing, 11, 1033–1046.
17. Lopez-Benitez, M., Umbert, A., & Casadevall, F. (2009). Evaluation of spectrum occupancy in
spain for cognitive radio applications. In IEEE 69th Vehicular Technology Conference, 2009.
VTC Spring (pp. 1–5).
18. Bao, V. N. Q., Cuong, L. Q., Phu, L. Q., Thuan, T. D., Quy, N. T., & Trung, L. M. (2011).
Vietnam spectrum occupancy measurements and analysis for cognitive radio applications. In
2011 International Conference on Advanced Technologies for Communications (ATC) (pp.
135–143).
19. Islam, M., Koh, C., Oh, S. W., Qing, X., Lai, Y., Wang, C., et al. (2008). Spectrum survey
in Singapore: Occupancy measurements and analyses. In 3rd International Conference on
Cognitive Radio Oriented Wireless Networks and Communications, 2008. CROWNCOM (pp.
1–7).
20. Cordeiro, C., Challapali, K., Birru, D., & Sai Shankar, N. (2005). IEEE 802.22: The first
worldwide wireless standard based on cognitive radios. In First IEEE International Symposium
on New Frontiers in Dynamic Spectrum Access Networks, 2005. DySPAN (pp. 328–337).
21. Mishra, V., Lau, C. T., & Chan, S. (2013). Reconfigurable channel selection technique for
cognitive radio network with heterogeneous primary bands. IEEE Transactions on Vehicular
Technology, 62, 3942–3952.
References 137
22. Song, Y., & Xie, J. (2010). Common hopping based proactive spectrum handoff in cognitive
radio ad hoc networks. In IEEE Global Telecommunications Conference (GLOBECOM 2010)
(pp. 1–5).
23. Lee, W.-Y., & Akyldiz, I. (2011). A spectrum decision framework for cognitive radio networks.
IEEE Transactions on Mobile Computing, 10, 161–174.
24. Garey, M. R., & Johnson, D. S. (1990). Computers and Intractability; A Guide to the Theory
of NP-Completeness. New York: W. H. Freeman & Co.
25. Working group on wireless regional area networks, IEEE Std. 802.22, 2011. http://www.
ieee802.org/22/.
26. Song, Y., Fang, Y., & Zhang, Y. (2007). Stochastic channel selection in cognitive radio networks.
In IEEE Global Telecommunications Conference, 2007. GLOBECOM’07 (pp. 4878–4882).
systems. In IEEE 21st International Symposium on Personal Indoor and Mobile Radio Com-
munications (PIMRC) (pp. 1831–1835).
28. Mishra, V., Tong, L. C., & Syin, C. (2012). Qos based spectrum decision framework for
cognitive radio networks. In 18th IEEE International Conference on Networks (ICON) (pp.
18–23).
Chapter 5
Case Study: Media Access in CRN
Framework
Based on the survey done in Sect. 1.7 and Chap. 2, the media access control proto-
cols are classified into two broad categories: random and time-slotted media access
schemes, each having its pros and cons. Recently a new class of access scheme, known
as hybrid schemes has been developed with better performance than that of the ran-
dom and time-slotted access schemes. Most of these scheme are network-model
dependent, and therefore a change in their operation from ad-hoc to infrastructure or
vice-versa is impractical. In a multi-channel environment, support for broadcast and
multicast is tedious in nature and requires implementation of flooding operation.
In this section, a media access scheme is discussed which is hybrid in nature
and provides unique features of multicast and broadcast without network flooding
[1]. The discussed media access scheme allows contention of nodes to be distributed
across channels, therefore mitigating the problems of channel saturation. The scheme
also allows nodes to prioritize packets for the purpose of QoS provisioning while
facilitating power saving mode for energy conservation.
As discussed in Sect. 3.2, the units responsible for the design of QoS provision-
ing media access scheme are highlighted in Fig. 5.1. The mac protocol unit (MPU)
receives the list of channels from channel selection and management unit (CSMU).
It also receives the information about the QoS requirement on long run statistical
behavior from the QoS control unit. Based on the information received the MPU
generates a MAC instance. The MAC instance on the basis of QoS requirement of
individual packet (access category) defines the individual packet behavior.
5.1 System Model
The system comprises of N cognitive radio user operating in an ad-hoc environment

comprising of C channels. A channel from one of the ISM bands or a channel with
high availability is selected as a rendezvous channel (RC). RC is utilized for the
140 5 Case Study: Media Access in CRN Framework

stack
Application
Application QoS
Engine
Routing Module
Energy Management
Control
Self
Coexistence
Engine

Database domain
MAC instance
MAC Protocol Unit
MAC instance
Layer 2
and above
Layer 1
Success/Failure
report
Sensing report
Fig. 5.1 Units in the CRN framework responsible for QoS provisioning media access scheme
purpose of network information interchange. The hardware complexity of CR users

is equivalent to that of the half-duplex transceiver with sensing capability.
5.1.1 Traffic Type
In modern world, conventional wireless network has to support multitude of different

data types due to increase in different types of applications. In a broad sense, data
traffic can be classified into four categories as discussed in Sect. 3.1.1 and shown in
Table 3.1. Some applications based on voice or video services require delay bounded
access, which require special handling of such packets.
As discussed in Sect. 3.1.1, the resource access delay suffered by a packet in the
network is governed by the access scheme utilized by the scheme. An access scheme
utilizes packet admission control through the introduction of different arbitration
spaces for different packets. Discussion on arbitration space is provided in Sect. 5.2.1.
5.1 System Model 141
Apart from arbitration space, QoS is governed by the reservation of bandwidth for a
particular traffic. This ensures that the traffic is provided with enough bandwidth to
support the level of QoS desired.
5.1.2 Channel Classification
The discussed scheme utilizes multiple channel for its operation, where different
channels serves different purposes. The channels utilized in the system are classi-
fied into two types: Data and Rendezvous channels. Users in the network perform
exchange of data packets over data channels while the exchange of control infor-
mation is done over rendezvous channel (RC). The discussed media access scheme
supports two types of data transfer: contention based and reservation based data
transfer. The support for dual data transfer types requires usage of two different
access etiquettes. Therefore data channels are further classified into Contention based
channels and Reservation based channels. Contention based channels (CBC) follow
the etiquette of contention based data transfer. Nodes which are on CBC perform
contention whenever a data transfer is required. Reservation based channels (RBC)
follow the etiquette of data transfer based on reservation. For the purpose of reser-
vation, nodes perform contention ahead of real data transfer on RBC. Therefore this
type of channel has two separate data period: contention (denoted by ad-hoc trans-
mission indication message (ATIM) period) and reserved data period. A reserved
data period is further sub-divided into data slots; each of which represents a data
packet transfer. During contention, nodes contend to reserve data slots, and during
reserved data period, scheduled nodes perform communication. Reservation based
channels allow implementation of higher priority stream based data services (for
example voice and video).
The system comprises of non-hopping nodes where nodes reside on a channel
as long as channel ceases to be free or nodes want to switch to another channel for
some reason. From a node’s perspective, channels are classified as Home channel
and Foreign channel. Home channel is a channel in which a node chooses to reside
while all other channels from this node’s perspective are termed as foreign channels.
A node belonging to home channel is termed as Home node and all other nodes are
termed as Foreign nodes.
5.1.3 Quiet Period Distribution
As discussed in Sect. 3.4.2, quiet period on a channel is a time duration during which
all CR users on that channel stop their communication so that effective spectrum
sensing can be performed. Quiet periods are designed to be periodic in nature, so
that nodes can detect an incumbent activity during the start of the current/next frame.
Sensing can be classified into two types: in-band and out-of-band sensing. In-band
sensing refers to sensing of home channel by home nodes while out-of-band sensing
refers to sensing done by nodes on foreign channels. To improve the overall through-
put of the system, the quiet periods on all channels are distributed such that they do
not overlap. This allows in-band and out-of-band sensing to be performed without
any throughput loss.
5.2 Hybrid Media Access Scheme
The design of the discussed MAC scheme with two reservation based channels,
two contention based channels and a rendezvous channel is shown in Fig. 5.2. Each
channel in the system is divided into fixed length time frames known as superframes.
The superframe of all data channels are overlapped and synchronized with each other.
Rendezvous channel is synchronized with data channels with a forward offset value of
size of rendezvous channel beacon (RCB). Therefore, the start of superframe of data
channels are overlapped and synchronized with the end of the RCB of rendezvous
channel.
Superframe
Rendezvous Contention based Data Transfer Rendezvous

Rendezvous
HCI
Channel Channel
QP
Channel Beacon Beacon
Superframe
Reserved Data Period
...
Reservation based
ATIM
S13
S14
S15
S16
S17
S18
S19
S10
S11
S12
S14
S15
S16
S17
S18
S19
HCI
S1
S2
S3
S4
S5
S6
S7
S8
S9
HCB
QP
channels
Data channels
...
ATIM
S13
S14
S15
S16
S17
S18
S19
S10
S11
S12
S14
S15
S16
S17
S18
S19
HCI
S1
S2
S3
S4
S5
S6
S7
S8
S9
HCB
QP
Contention based Data Transfer
...
Contention based
HCI
HCB
QP
channels
...
HCI
HCB
QP
ATIM
Ad-Hoc Transfer
QP
Quiet Period (Sensing)

HCB Home Channel Beacon Indication Message
Home Channel Xth mini time-slot in

Sx
HCI
Information Beacon reserved data period
Fig. 5.2 MAC frame structure

5.2 Hybrid Media Access Scheme 143
The superframe of all channels is composed of different intervals designated for

specific purposes. Common intervals among all superframes are quiet periods (QP)
and home channel information beacon (HCI). During QP, all nodes on that channel
stop their transmission and perform in-band sensing. During this period, any foreign
node (i.e. node on some different channel than that of the current QP) can perform
out-of-band sensing on this channel. During HCI, all nodes in their respective home
channels compete to select a representative node (RN). A representative node is
chosen periodically from each channel which broadcasts information about its home
channel onto rendezvous channel. The information consists of number of nodes on
home channel, quiet period distribution and channel characteristics. Apart from QP
and HCI, rendezvous channel comprises of rendezvous channel beacon (RCB) and
contention based data transfer period. A RCB has a time slotted design, with total
number of slots designated as t. During RCB, representative nodes (RN) switch to
rendezvous channel and share information about their home channel. Each slot in
RCB is designated to each RN, therefore t = C + 2; where 2 additional slots are
required for new nodes joining the network.
Apart from QP and HCI, the superframe of a data channel comprises of home
channel beacon (HCB) and data transfer period. During HCB, the RN returns from
rendezvous channel to its home channel, and broadcasts the information it learned
from other representative nodes in the network. A HCB on a channel has a time
slotted architecture, where number of time-slots equals the number of nodes on that
channel +1. An extra time slot is provided for a new node joining the network on
that channel. Depending upon data channel type, the data transfer can be classified
into two types: reservation and contention based data transfer. A reservation based
channel (RBC) employs reservation based data transfer service, while a contention
based channel (CBC) employs contention based data transfer service. RBC consists
of ATIM (ad-hoc transfer indication message) period and reserve data slots. Dur-
ing ATIM period, communicating node performs contention, and reserve data slots
ahead of actual data transmission. A three way handshake is performed during the
reservation process. The reservation of a resource block before actual communi-
cation, allows implementation of higher priority service. The design of contiguous
resource blocks allows provisioning of stream based services like voice or video
which require contiguous stream of data transfer [2]. On the other hand contention
data period on CBC requires nodes to perform a 4 way handshake similar to optional
RTS/CTS mechanism of 802.11 [3]. Therefore nodes contend whenever they require
to perform a data transfer during the contention data period. Contention based data
transfer service is also followed by nodes on rendezvous channel.
5.2.1 Arbitration Interframe Spaces
For the purpose of packet prioritization, different arbitration interframe spaces are
utilized. Arbitration interframe space is the time duration which a node has to wait
before transmitting a packet. Arbitration in this discussed scheme is different from
Table 5.1 Interframe spaces for different packet types

Interframe space RTS packet type/Other IFSmin IFSmxo
packet type
SIFS Voice/CTS/ACK/DATA IFSsmin IFSsmax
MIFS Video IFSmmin IFSmmax
IIFS Best effort IFSimin IFSimax
LIFS Background data IFSlmin IFSlmax
the traditional arbitration as in enhanced distributed channel access (EDCA) [4],

where a fixed value is utilized and the value of contention window is changed on
each retransmission. Here arbitration is defined as a random duration in [IFSmin ,
IFSmax ]. Here IFSmax = 2r .IFSmxo , where r is the number of retransmission per-
formed for this packet, and IFSmxo is the initial maximal boundary for arbitration
value selection. Each packet before transmission is assigned a counter which is ini-
tialized with the value of arbitration space. The counter is decremented whenever the
channel is sensed to be free. If the counter reaches zero, the packet is transmitted.
If transmission is unsuccessful, then the value of r is incremented and the packet is
scheduled for retransmission. If the packet is not deliverable even after number of
retransmission, it is dropped. The value is called as retry limit. Different limits of
arbitration interframe space for different types of packets are shown in Table 5.1, in
which S/M/I/L-IFS implies short, medium, intermediate and large interframe space
respectively. Therefore an RTS packet of video has to wait MIFS duration before
transmission. Similarly ACK packet of any data type has to wait for SIFS duration
before transmission.
5.3 Initialization
The cognitive radio network initializes by selecting a rendezvous channel. A node

in the network scans for a RCB while performing sensing on all defined channels.
If a node does not detect any RCB, then it starts its own RCB by first selecting
an appropriate channel as a rendezvous channel (RC). An auxiliary RC may also
be selected as a backup, just in case main RC ceases to be free. If a node detects
an existing RCB, it joins through by doing transmission on the extra time-slots of
RCB. Based on the information it learned from RCB, it may join a data channel by
following the same procedure on HCB. A node having high probability of performing
a high priority data transmission will select reservation based channel. If there does
not exist a data channel, the node may start its own beacon for the data channel.
The node may classify a data channel as RBC or CBC depending upon channel
availability. A channel having high availability is considered for the candidate of
RBC and vice-versa.
5.4 Operation 145
5.4 Operation
Since two different types of data channels are utilized in the access scheme, two dif-
ferent etiquettes of channel access are also utilized. The nodes follow the etiquette
of the channel in which they perform communication. The communication may take
place on either of the home channels of communicating nodes. If the desired char-
acteristics of communication media is not supported by either home channels of the
communicating nodes, then the communication may take place on a foreign channel.
In this case the etiquettes of the foreign channel will be utilized for communication
purposes. Depending upon the channel characteristic and location (home channel)
of sender and receiver, three different types of data transfer are classified:
5.4.1 Data Transfer on Reservation Based Channels
On a reservation based channel, nodes contend to reserve data slots during ATIM
period. For contending, nodes employ three way handshake to reserve data slots
ahead of actual communication. A node initiates this by sending first request-to-send
(RTS) packet. The RTS packet of sender consists of a list identifying available data
time-slots onto which the sender intends to communicate. The receiver replies with
clear-to-send (CTS) packet containing the list of data time-slots on which it agrees
to receive the data. The sender confirms this list by sending an acknowledgement
(ACK). All other nodes know which slots are reserved for what type of data transfer.
A high priority data transfer like voice or video can preempt the reserved data slots
from scheduled low priority data transfer service; and therefore high priority service
can schedule its own transmission in those reserved slots as if they are not reserved.
During the reserved data period, scheduled nodes perform data communication.
5.4.2 Data Transfer on Contention Based Channels
Data transfer on a contention based channel follows the traditional optional 4 way
handshake RTS/CTS mechanism of 802.11 [5]. A sender node initiates the transmis-
sion by sending an RTS packet containing the duration of the transmission in a field
called as network allocation vector (NAV). The receiver replies with CTS containing
NAV, to identify that it is ready to receive the data. The sender then performs trans-
mission of DATA packets. After the reception of DATA packet, the receiver replies
with an acknowledgement (ACK) packet.
5.4.3 Data Transfer on a Foreign Channel
If a sender has a packet for a receiver which does not lie on the home channel
of the sender, the sender moves to the home channel of receiver after HCB. The
sender then performs communication with the receiver using the channel etiquettes
of the receiver. Once the communication is finished or HCI of the home channel of
the sender starts, the sender moves back to its own home channel. The procedure is
repeated until the receiver node decides to move onto the home channel of the sender.
If the desired characteristic of the channel is not supported by the home chan-
nel of receiver, then data transmission is performed on a foreign channel with the
desired characteristics. The sender moves to the home channel of the receiver and
sends a MOV packet to the receiver indicating the channel into which they should
switch onto. If the receiver agrees, it replies with an acknowledgement MOV-ACK
and immediately switches to the indicated channel. The sender after receiving the
acknowledgement also switches to the indicated channel. Both sender and receiver
node waits for LIFS duration on the indicated channel to overhear any current com-
munication and then follow their etiquettes. Both sender and receiver move back to
their home channel when the communication is finished or HCI of the home channels
of sender/receiver starts.
5.5 Power Saving Mode Operation
The discussed MAC scheme allows different power saving mode operation for nodes
depending upon their home channel. Nodes are always active during the period of
beacon exchange (RCB, HCB, HCI) on their home channel. Here active duration
refers to the time duration for transmit, receive or sensing operation.
On a reservation based channel, nodes are always active during HCB, ATIM,
QP and HCI period. A node learns about the usage of channel (data slots) during
ATIM period, and therefore schedules to enter into doze (idle) mode if it is not
performing any communication. The node wakes up when it has its turn for sending
or receiving data on reserved data time-slots. Figure 5.3 shows 3 nodes A, B and C
operating on a reservation based channel. During ATIM period, nodes A and B have
scheduled communication on data time-slots S5 − S14. At the end of ATIM, nodes
A, B and C enter into doze mode. At the start of data slot S5, node A and B become
active and perform communication. During QP, node A and B stop their transmission
and perform sensing operation. During QP, node C wakes up and performs sensing
operation. During HCI, all nodes exchange their beacon for the purpose of selection
of a RN. Since node A and B are busy in performing data communication, some
other node is chosen as RN. Once the communication is done, node A and B again
enter into doze mode.
The power saving operation on contention based channel is followed similar to
the power saving mode (PSM) of 802.11 [3]. All operating nodes on the channel
5.5 Power Saving Mode Operation 147
ith Superframe
Active
Duration
...
ATIM
S13
S14
S15
S16
S17
S18
S19
S10
S11
S12
S14
S15
S16
S17
S18
S19
HCI
Node A
S1
S2
S3
S4
S5
S6
S7
S8
S9
HCB
QP
exchange
packets
ATIM
Active
Duration
...
ATIM
S13
S14
S15
S16
S17
S18
S19
S10
S11
S12
S14
S15
S16
S17
S18
S19
HCI
S1
S2
S3
S4
S5
S6
S7
S8
S9
Node B HCB
QP
Active
Duration
...
ATIM
S13
S14
S15
S16
S17
S18
S19
S10
S11
S12
S14
S15
S16
S17
S18
S19
HCI
S1
S2
S3
S4
S5
S6
S7
S8
S9
Node C HCB
QP
Fig. 5.3 Power saving mode operation on a reservation based channel of the MAC scheme
overhear the RTS and CTS packet of communicating nodes, which consists of a
network allocation vector (NAV). Since NAV indicates the duration of transmission,
all other nodes can calculate the time for which they should remain in doze mode.
Therefore nodes sleep during an ongoing transmission of other nodes, and wake up
when the transmission is finished. The value of NAV for RTS and CTS packet is
calculated as:
NAVRTS = tCTS + tDATA + tACK
NAVCTS = tDATA + tACK
where tpkt is the time duration required for the transmission of a packet pkt.
5.6 Broadcast and Multicast Operation
Broadcast and multicast operation are required by many higher/lower layer protocols
to implement certain functionality. Broadcast operation in multi-channel environ-
ment faces the problem of dissipation of a single packet over multiple channels and
hence multiple channel hopping is required. The discussed scheme mitigates such
a problem by providing a broadcast or multicast operation within 2 superframe. A
node which has a packet to broadcast will select itself as a representative node and
broadcast the packet in RCB of next superframe. The representative nodes from other
channels on returning to their home channel from RCB will broadcast the packet in
HCB, which allow all nodes on that channel to receive the broadcasted packet; since
this follows on all channels, the broadcasted packet are received by all the nodes in
the network. Multicast operation follows similar to the broadcast operation, except
that a masking operation is required, which depends upon multicast application.
5.7 Performance Analysis
The discussed scheme is evaluated using a discrete event simulation in Matlab which
simulates the network up to frame level. The channel availability model of Two-state
Markov chain from Sect. 3.4.1 is utilized, with the channel state-transition probabili-
ties, α and β set to 0.9 and 0.1 respectively. To clearly show the distinct performance
on different data channel (RBC and CBC), the performance of each channel is indi-
vidually evaluated. Users can make a decision on the basis of latency and throughput
requirement to select an appropriate channel. Four different type of data packets are
utilized: voice, video, best effort and background data. Two reservation based channel
and two contention based channels are evaluated with 4 users on each channel.
On each channel, two senders are randomly selected from 4 users, while rest
two as destination. The arrival rate of data packets on these users is modeled as
Poisson arrival with the percentage distribution of aggregate arrival as voice, video,
best effort and background data packets set as 10, 20, 60 and 10 % respectively. The
parameters used for the simulation are shown in Table 5.2. To simplify the analysis,
bandwidth of each channels is assumed to be equal and the size of RTS, CTS and
ACK packets used on RBC and CBC are assumed to be equal. The data packet
size of all 4 different data types are also assumed to be equal. To capture the effect
of stream based transfer on RBC, a transmission of 3 voice packets requires only
a single RTS/CTS/ACK handshake exchange during ATIM duration; similarly a
transmission of 4 video packets requires only a single RTS/CTS/ACK handshake
exchange during ATIM duration. While all other transmissions require exchange of
RTS/CTS/ACK for every data packet. Similarly on CBC, all transmissions require
exchange of RTS/CTS/ACK for every data packet. The arrival rate for each user is
increased from 50 packets per second to 330 packets per second with step size of 20.
The simulation time is 100 s, and each data point in the graph is the average of 100
run.
In Fig. 5.4, analysis of average latency of individual (in terms of priority) packets
is done. A clear comparison of Fig. 5.4a, b evaluates that, packets on RBC has a
low and defined (calculable) latency in comparison to CBC. Moreover the latency
for voice and video packets in RBC does not increase as they tend to increase in
CBC with increase in arrival rate. A calculable latency on RBC is attributed to its
reservation based design, which allows data slots to be reserved ahead of actual
communication. The main drawback of RBC etiquette in comparison to CBC is that,
it has a high latency for low level priority packets.
Figure 5.5 shows the probability of success of transmission of each originating
packet. From Fig. 5.5a, b, it can be concluded that as the arrival rate increases, there is
a decrease in success probability due to more number of collisions. Also the success
ratio of RBC for higher priority packets (voice and video) is more in comparison to
CBC. This can be attributed to pre-scheduled DATA packets in RBC, where as in
CBC, large DATA packets result in backoff or collision of other packets. The success
ratio of low priority (best effort and background) data packets is lower in RBC due
5.7 Performance Analysis 149
Table 5.2 Simulation Parameters Value

parameters
Simulation time 100 s
Slot time 10 µs
Number of RBC 2
Number of CBC 2
Bandwidth of channels 2 Mbps
Superframe size 100 ms
RCB beacon size 15 ms
HCB beacon size 3 ms
HCI beacon size 2 ms
Quiet period size 2.5 ms
ATIM window size 35 ms
Number of reserved data slots 24
Reserved data time slot size 2.5 ms
RTS packet size 240 bits
CTS packet size 240 bits
ACK packet size 240 bits
DATA packet size 5000 bits
IFSsmin , IFSmmin 1
IFSsmin , IFSmmin 1
IFSimin , IFSlmin 3
IFSsmax 3
IFSmmax 5
IFSimax 7
IFSlmax 10
5
to the preemptive nature of high priority packets to take away the communication
segments (data slots) reserved for low priority packets.
Figure 5.6 shows the channel utilization of RBC and CBC. Here, the channel
utilization is calculated on the basis of actual DATA packets transmitted. Therefore,
the channel utilization does not include packets like RTS, CTS or ACK. At low
arrival rate, RBC tends to perform better due to low collision in ATIM contention
window. The lower collision is attributed to small size of control packets, where
actual large DATA packets are communicated during reserved data slots. But after
a threshold (aggregate number of arrivals per node = 170), CBC performs better in
terms of channel utilization. This is attributed to restricted contention window size
in RBC. The small ATIM contention window in RBC cannot accommodate more
than a defined number of control packets, and hence actual communication is limited
by it. Therefore, CBC achieves a higher channel utilization due to large contention
based data transfer period.
0.08
0.1 Voice data

0.07
Video data
0.09
Best effort data
0.06
0.08 Background data
Latency (seconds)
0.07
Latency (seconds)
0.05
Voice data
0.06 Video data
Best effort data 0.04
0.05 Background data
0.04 0.03
0.03
0.02
0.02
0.01
0.01
0 0
50 100 150 200 250 300 350 50 100 150 200 250 300 350
Aggregate arrival rate per node Aggregate arrival rate per node
Fig. 5.4 Latency of packets on different channels
0.8
Success ratio
0.6 Voice data

Video data
Background data
0.2
50 100 150 200 250 300 350
Aggregate arrival per node
0.95
0.9
Success ratio
0.85
0.8 Voice data
0.75 Video data
Background data
0.65
50 100 150 200 250 300 350

Fig. 5.5 Success ratio of packets on different channels

5.8 Infrastructure Mode Operation 151
0.8
0.7
Utilization 0.6
0.5
0.4
0.3 Reservation based channel

Contention based channel
0.2
0.1
50 100 150 200 250 300 350
Fig. 5.6 Channel utilization
5.8 Infrastructure Mode Operation
The discussed scheme can also be employed in infrastructure mode by doing slight
modification in middleware without any change on hardware requirements of users
node. For operation in infrastructure mode, network employs an access point (AP)
with the number of transceivers equal to the number of data channels employed.
Data channels are formulated by AP with quiet periods distributed in a predefined
sequence. The requirement of a rendezvous channel is removed and the AP acts as a
gateway to all the nodes operating in the network. Access point acts as representative
node (RN) on all data channels and the requirement of home channel information
(HCI) beacon is also removed. Therefore all data channels (including RBC and
CBC) are synchronized, and overlap with each other. During HCB, AP learns about
the number of active nodes on a channel, channel characteristics and information
regarding channel sensing from different nodes to perform cooperative sensing. Any
node that wants to send a packet to another node, passes it through access point.
Therefore user nodes do not hop for packet transmission. An AP also calculates
and broadcasts average latency of packets on a given channel, so that a node can
join/move to a given channel depending upon the observed latency. Nodes join/leave
a channel through the help of AP; while AP also coordinates in assignment/removal
of beacon slot for users node in HCB.
The infrastructure mode, operation of the discussed MAC scheme is shown in
Fig. 5.7, where different equipments in home/office with different priority services
operate with each other. An access point acts as a gateway and allows Internet con-
nection through a switch. High priority services like IP phone and smartphone will
operate with voice/video priority services. IP TV and media servers will operate with
Fig. 5.7 Infrastructure mode

operation with network
equipments
Media Server IP phone
IP TV
Printer
Access Point
Desktop
Smartphone
Switch
PDA
File Server Laptop
video priority but laptops and desktop will operate with hybrid priority (video and
best effort) depending upon latency requirement, file servers and PDA will operate
with best effort priority; while printers will operate with background priority.
In infrastructure mode, all nodes are connected to the AP, and hence are not hidden
from it. Nodes may be hidden from each other, due to location and transmission power
constraints. Since nodes perform three way or four way data exchange (depending
upon channel type), the hidden terminal problem is solved.
5.9 Hidden Terminal Problem in Ad-hoc Mode
The discussed access scheme in ad-hoc mode is only defined for a network consisting
of single-hop nodes. In multi-hop scenario, it may be possible that the information
broadcasted by a rendezvous node (RN) is not be heard by some nodes on that
channel due to the difference in location. Similarly, the information transmitted by
some nodes on home channel beacon may not be heard by other home nodes. This
is called hidden terminal problem, which results in nodes having inconsistent set
of information. Multiple solutions exist to solve this problem. One solution is to
increase the size of home beacon and ask every home node to repeat the information
it learned from RN and other home nodes; therefore the information broadcasted
is cumulative in nature. A node which is hidden to RN or to any other node, will
overhear it and thus they all will have consistent set of information. Another solution
is that all nodes generate its next-hop neighbor list. A next hop neighbor list of a node
consists of list of neighbor nodes that the node learns from overhearing packets on
the channel. This next hop neighbor list is matched with the information broadcast
on HCB. If there are inconsistencies among the list of visible nodes on the channel,
they are resolved.
5.10 Discussion 153
5.10 Discussion
In a cognitive radio network, multiple cognitive nodes try to access sporadically

varying resources (frequency spectrum) which results in access collision. To mitigate
access collision, an access scheme is required which is followed by all the cognitive
nodes operating in the cognitive radio network. The etiquettes of an access scheme
is directly responsible for quality of service from latency perspective, while it is
also responsible for quality of service from throughput perspective to an extent. The
queuing delay which directly affects latency is controlled by the access scheme while
throughput is directly controlled by the channel selection scheme and indirectly by
the design of media access scheme.
In this chapter, different media access schemes for cognitive radio networks were
discussed and a case study was performed on a QoS aware hybrid media access
scheme framework for cognitive radio network. The scheme utilizes a hybrid of
two different channel etiquettes: reservation based channel and contention based
channel. These channels have different attributes and hence support different types
of services. The discussed media access scheme utilizes power saving mode for
doze mode operation while supporting major network operations like broadcast and
multicast. An infrastructure based operation is also discussed which requires slight
modification on middleware, while no changes are required on hardware of user node.
Hidden terminal problems in infrastructure mode and ad-hoc mode are also discussed.
Performance analysis of the discussed scheme shows that reservation based channels
have low latency and higher transmission success probability for higher priority
services. In contrast, the channel utilization of contention based channel achieves
better performance than that of the reservation based channel for higher arrival rate.
Even though a media access scheme mitigates access collisions within a network, it
cannot directly mitigate access collision with another network.
References
1. Mishra, V., Tong, L.C., Syin, C. (2013) A qos provisioning mac protocol for cognitive radio net-
work. In 2013 7th International Conference on Signal Processing and Communication Systems
(ICSPCS) (pp. 1–8).
2. Mishra, V., Tong, L.C., Chan, S., Mathew, J. Mac protocol for two level qos support in cognitive
radio network. In 2011 International Symposium on Electronic System Design (ISED) (pp. 296–
301).
3. IEEE 802.11 working group, wireless lan medium access control (mac) and physical layer (phy)
specifications. IEEE 802.11 Standard (2012).
4. Perez-Costa, X., & Camps-Mur, D. (2010). IEEE 802.11e qos and power saving features
overview and analysis of combined performance [accepted from open call]. IEEE Wireless
Communication, 17, 88–96.
5. Sakurai, T., & Vu, H. (2007). Mac access delay of ieee 802.11 dcf. IEEE Transactions on Wireless
Communication, 6, 1702–1710.
Chapter 6
Case Study: Energy Management in CRN
Framework
In Chap. 2, different spectrum selection strategies were discussed for the purpose
of QoS provisioning in deterministic and stochastic channel environment. In this
chapter, an energy aware spectrum decision scheme for QoS provisioning CR frame-
work (ESDF-CR) is discussed in which a user selects a channel which reduces the
power consumption of transmission [1]. The scheme kicks in when the remaining
energy of the user falls below a given threshold, so as to allow extended operation
of the system with graceful degradation in QoS. The selected channel has lower
bandwidth and thus has limited throughput capacity for transmission which results
in queuing of packets. A packet admission control scheme is designed so as to allow
high priority packets to pass through the selected channel while blocking the lower
priority packets. Thus a graceful degradation of QoS can be maintained by the system.
The technologies/modules responsible for this energy aware spectrum decision are
highlighted in the discussed QoS provisioning framework as shown in Fig. 6.1. The
energy management unit (EMU) provides the information and constraints regarding
the remaining power of the system, whether it is running on AC power or battery.
When the remaining power of the system fall below a certain threshold, the EMU asks
for energy aware spectrum decision through channel selection and management unit
(CSMU). The EMU also gets the information regarding QoS from QoS control unit
(QCU) and perform packet admission control accordingly. This framework extend
the energy management feature by allowing QoS control unit to communicate with
energy management unit so as to control the transmission power. This allows Layer
1 developers of the protocol stack to easily incorporate the design of transmission
power control in a cognitive radio user.

156 6 Case Study: Energy Management in CRN Framework

stack
Application
Application QoS
Engine
Routing Module
Energy Management
Control
Self
Coexistence
Engine

Database domain
MAC instance
MAC Protocol Unit
MAC instance
Layer 2
and above
Layer 1
Success/Failure
report
Sensing report
Fig. 6.1 Modules responsible for energy aware spectrum decision in the discussed QoS provision-
ing framework
6.1 System Model
The system comprises of a single cognitive radio network in ad-hoc mode, comprising
N CR users each of which is equipped with a single half-duplex transceiver. The
system operates in a heterogeneous environment comprising of C channels, each
having bandwidth Bi , ∀i ∈ {1, ..., C}. The transmit power required on each channel
is assumed to be proportional to the bandwidth of the channel and it is represented by
Pi , ∀i ∈ {1, ..., C}. The spectral opportunity is homogeneous to all the CR users of
the networks in terms of availability. Each of these channels is modeled as two state
ON-OFF Markov chain as discussed in Sect. 3.4.1. Each CR node is also assumed
to have a limited battery capacity. All CR users keep track of the available channels
with an ideal channel sensing scheme. For the purpose of provisioning of QoS, the
system classifies the user traffic into four different classes as discussed in Sect. 3.1.1.
6.2 Energy Aware Spectrum Allocation Scheme 157
6.2 Energy Aware Spectrum Allocation Scheme
The main assumption of the discussed scheme is that a smaller bandwidth channel
will consume lower power for communication than its counterpart. The overall work
of energy aware spectrum decision can be divided into two parts [2]:
1. Selecting a channel which utilizes low power.
2. Admission control on low priority packets.
The system first selects a channel satisfying the bare bandwidth need of the important
applications (having higher priority). Once the bare bandwidth requirement is satis-
fied, the system performs admission control to the rest of the applications packets to
utilize the rest of the bandwidth of the channel. The energy management unit (EMU)
executes the threshold check for invoking energy aware spectrum allocation scheme
in channel selection and management unit (CSMU). On receiving this command, the
CSMU gets the information of bare bandwidth demand (discussed later) from QoS
control unit (QCU) and then perform spectrum allocation accordingly which satis-
fies this bare bandwidth. Even with the allocation of the selected channel by higher
priority packets, there is some bandwidth left on the channel through which lower
priority packets can be transmitted. The packet admission control unit (PAC) allow
scheduling and queuing of lower priority packets through remaining bandwidth of
the channel.
6.2.1 Spectrum Decision
The energy management unit (EMU) keeps track of the power level and kicks in
the energy aware spectrum decision at the channel selection and management unit
(CSMU) when the power level drops below a certain threshold (user selectable). A
guarantee on QoS in terms of bandwidth is provided to the top two applications:
Voice and Video (from Table 3.1). Algorithm 6.1 provides the energy aware channel
selection scheme which executes at the CSMU.
The channel indices used in Algorithm 6.1 is sorted in the order of bandwidth.
Here bare bandwidth (ψ) is defined as the minimum bandwidth required to provide
assured grade of QoS to the top two applications (voice and video packets). Therefore
it can be assumed that the bare bandwidth is the bandwidth requirement of the
top two applications and it can change depending upon the environment. A search
is performed starting from lower bandwidth channels to find a channel which is
available and has bandwidth greater than bare bandwidth demand i.e. Bi ≥ ψ. The
band() function returns the bandwidth of the parameter passed. Here voice and
video parameter represent the total number of voice and video packets respectively
queued in the given time frame. To identify whether a channel is available or not,
function I s Fr ee() is used. The function returns a value 1 when the channel ID passed
is available otherwise it returns a value 0. If no channel satisfies the need (i.e. the
Algorithm 6.1: Algorithm for selecting a channel at a given time frame

1 Initialize : ψ = band(voice) + band(video);
2 Label:
3 for i = 1 to C do
4 if band(channel(i)) ≥ ψ then
5 if I s Fr ee(channel(i)) then
6 select_channel = i;
7 br eak − f or ;
8 end
9 end
10 end
11 if select_channel = φ then
12 ψ = band(voice);
13 goto Label;
14 end
bandwidth ψ), then the bare bandwidth requirement is reduced to only voice packets.
If even after this no channel is found to satisfy the requirement, then the currently
selected channel is used to transmit low priority packets.
6.2.2 Bare Bandwidth Calculation
As discussed previously, bare bandwidth is calculated as sum of bandwidth require-

ment by voice and video packets. Within voice and video packets it is possible to
have different access levels based on the waiting time of the packets in the queue.
This can be followed by the fact that a video conference application has strict latency
requirement for video packets than media applications like Youtube where packets
can buffer if needed. Therefore to minimize the bandwidth use by different applica-
tions, the number of packet scheduling is reduced in such a way that packets with
large latencies are scheduled first. Each application generating real time packet is
assigned to have a maximum queue latency that it can suffer before performance
degradation. This queue latency is given as δi for packets generated by video and
voice applications, where i represents a packet ID. Ai and T represent the arrival
time of packet i and current system time respectively. A packet to be considered for
transmission in the current frame is selected using Algorithm 6.2 which executes
at the QoS control unit (QCU) and is managed with the help of packet admission
control unit (PAC).
The algorithm calculates the current latency of all the packets and selects only
those which satisfy the given constraint in line 4. The expression (T − A pkt ) gives
the current waiting time of the given packet ( pkt). The expression (T − A pkt +
FrameDuration) gives the maximum latency which the given packet will expect if
it is deferred from transmission in the current time frame. Therefore the condition in
6.2 Energy Aware Spectrum Allocation Scheme 159
Algorithm 6.2: Algorithm for bare bandwidth calculation at a given time frame
1 Initialize : select_pkts = φ;
2 for i = 1 to Si ze(Q) do
3 pkt = Q i ;
4 if (T − A pkt + FrameDuration) > δ pkt then
5 select_pkts = select_pkts ∪ { pkt};
6 end
7 end
8 ψ = band(select_pkts);
line 4 evaluates to true for packets which must be transmitted in the current time frame
to meet the strict latency requirement and alleviate the system from performance
degradation.
6.2.3 Admission Control
Once the channel is selected satisfying the bare bandwidth need, there is still some
bandwidth left in the channel to allow the lower priority packets to pass through.
An admission control technique is shown in Algorithm 6.3 to allow these packets
to utilize the remaining bandwidth. This function is implemented in packet admis-
sion control unit of the framework. Even after transmission through the remaining
bandwidth, some of the low priority packets are still blocked and hence need to be
scheduled for transmission in next time frame. The discussed algorithm considers
this scenario and attaches a counter to each blocked packet. The counter is initialized
to zer o when the packet is blocked for the first time and is then incremented by a value
of 1 whenever the packet is not sent in the next successive time frame. Whenever a
packet selection is done, a packet having largest counter value is selected.
Algorithm 6.3: Algorithm for admission control to low priority packets in a

given time frame
1 Initialize : λ = band(select_channel) − ψ;
2 num_best_ pkt = (γ ∗ λ)/(best_ pkt_si ze);
3 num_backgr ound_ pkt = ((1 − γ) ∗ λ)/(backgr ound_ pkt_si ze);
4
5 select_best_pkts = num_best_ pkt f r om highest_counter (ζ);
6 select_background_pkts = num_backgr ound_ pkt f r om highest_counter (ξ);
7 ζ = ζ − {select_best_ pkts};
8 ξ = ξ − {select_backgr ound_ pkts};
9 counter (ζ) = counter (ζ) + 1;
10 counter (ξ) = counter (ξ) + 1;
An assumption is made that the selected channel already satisfies the bare band-
width requirement for all voice and video packets. Therefore low priority data like
best effort data and background data (Table 3.1) can share the remaining bandwidth
in a definite way. It can be assumed that the best effort data takes a share γ (0
≤ γ ≤ 1) of the remaining bandwidth while background data takes (1 − γ) share
of the remaining bandwidth (λ). The algorithm first calculates the remaining band-
width λ of the selected channel. The number of best effort packets and background
packets are then calculated which can be allowed to pass through the channel (lines
2–3). The algorithm then selects these packets from the appropriate category having
higher counter values compared to other blocked packets (lines 5–6). The set ζ rep-
resents the set of blocked best effort packets while set ξ represents the set of blocked
background data packets (lines 7–8). The function counter () is used to update the
counter value of each packet in the set (lines 9–10). The set select_best_ pkts and
select_backgr ound_ pkts consists of the packets allowed to pass through the current
time frame.
6.3 Performance Analysis
The discussed spectrum decision scheme is compared against random channel selec-
tion scheme [3] due to lack of other comparable spectrum selection schemes with
similar attributes. A random channel selection scheme is a channel selection tech-
nique in which a free channel is selected randomly and is used until that channel
is sensed occupied by the primary user [3]. For a fair comparison, it is assumed
that the random channel selection scheme selects a channel which satisfies the bare
bandwidth demand of the node. The discussed scheme is evaluated using discrete
event simulation in Matlab which simulates the network up to frame level. The media
access scheme utilized for the performance evaluation is taken from Chap. 5. All the
nodes in the network are assumed to be at one-hop distance to each other. The per-
formance of the discussed scheme is evaluated in terms of energy consumption of
the system and number of low priority blocked packets. The primary channels in
the system are modeled as two state Markov chain as discussed in Sect. 3.4.1 with
channel transition parameters α and β assumed to be uniformly distributed over
the range [0, 1]. The bandwidth of the channel is modelled linear with the num-
ber of channels used, i.e. the minimum bandwidth of a channel is given a value of
0.5 Mbps and the maximum bandwidth of the channel is 0.5 ∗ C Mbps, where C is
the maximum number of channels used for the simulation. Therefore the bandwidth
is distributed linearly from channel 1 to channel C as 0.5 to 0.5 ∗ C Mbps. Similarly,
the transmit power level on these channels is modeled to be linear with the number
of channels used and its value varies from 0.1 to 0.1 ∗ C watts. The number of voice
communication packets, video communication packets, best-effort communication
packets and background traffic communication packets used per second are 1, 2, 15
and 7 respectively giving the percentage ratio of 4, 10, 60 and 26 % for voice, video,
best-effort and background traffic communication segments respectively. The size
of voice, video, best-effort and background packets are 0.05, 0.5, 0.05 and 0.05 Mb
respectively.
A relation between energy consumption and simulation time of the system is
shown in Fig. 6.2. The number of primary channels used is 20 (C = 20). The dis-
cussed spectrum selection scheme (ESDF-CR) consumes 50 % less energy than ran-
dom channel selection scheme. The result follows from the selection of channels
which require low power for transmission and have satisfactory bandwidth. Thus the
discussed energy aware spectrum decision scheme increases the operable life of the
system.
A comparison between different low priority blocked data packets with respect to
simulation time of the system is shown in Fig. 6.3. The number of primary channels
used is 20 (C = 20). The total number of blocked best effort data packet and
background data packet for ESDF-CR scheme increases faster than random channel
selection scheme. This could be followed from the results in Fig. 6.2, where the
discussed scheme selects low transmission power channels and thus utilizes lower
bandwidth in comparison to that in the random channel selection scheme. The usage
of low bandwidth channels results in blocking of low priority data packets.
Fig. 6.2 Relation between 900
energy consumption in the 800

system with simulation time ESDF−CR
700 Random channel selection
Energy consumed (joules)
600
500
400
300
200
100
0
0 100 200 300 400 500 600 700 800 900 1000
Simulation time (seconds)
Fig. 6.3 Relation between 8000
number of packets blocked best effort data through ESDF−CR

7000 background data through ESDF−CR
with simulation time best effort data through random channel selection
6000 background data through random channel selection
Total no. of packets blocked
5000
4000
3000
2000
1000
0
0 100 200 300 400 500 600 700 800 900 1000
Simulation time (seconds)
A relation between energy consumption of the system with the size of voice packet
used by the application is shown in Fig. 6.4. The size of voice packet is varied from
0.05 to 0.5 Mb with the 20 primary channels (C = 20) and 100 s of simulation time.
The size of other data traffic packets is kept unchanged. As the size of the voice
packet increases the energy consumption also increases for ESDF-CR which can be
followed from the requirement of higher bandwidth channels which consumes more
energy. The effect of the variation of voice packet size can also be seen in random
channel selection scheme where it tends to randomly select a channel which satisfies
the bandwidth demand of varying size voice packet.
Figure 6.5 shows the relation between the number of blocked packets in the net-
work with the size of voice packet used by the application for 20 primary channels
(C = 20) and 100 s of simulation time. The size of voice packet is varied from 0.05
to 0.5 Mb. As the size of the voice packet increases the bare bandwidth demand also
increases which results in selection of high bandwidth channels and higher energy
consumption (Fig. 6.4). Since the bandwidth of channels used varies linearly with
Fig. 6.4 Energy 90
consumption versus size of

voice packet 80
70
60 ESDF−CR
Random channel selection
50
40
30
20
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Size of voice packet (Mb)
Fig. 6.5 Number of packets 800
blocked versus size of voice

700
packet
600
best effort data through ESDF−CR

500 background data through ESDF−CR
best effort data through random channel sel.
400 background data through random channel sel.
300
200
100
0
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Size of voice packet (Mb)
Fig. 6.6 Energy consumed 110
versus total number of 100

communication segments
initiated 90

80
70 ESDF−CR
Random channel selection
60
50
40
30
20
0 0.5 1 1.5 2 2.5 3 3.5 4
4
Total no. of communication segments initiated x 10
number of channels, and the number of voice segments are discrete in nature, a higher
bandwidth channel will result in more unoccupied bandwidth after a certain thresh-
old. Thus more best effort data and background data are allowed to pass resulting in
decreased blocking rate.
A relation between total energy consumed in the system with the total number of
data segments generated per 100 s is shown in Fig. 6.6. The size of voice, video, best
effort and background traffic packets are all set to 0.05 Mb with 20 primary channels
used in the system. The energy consumed by ESDF-CR increases with the number of
communication segments used in the network. Also the discussed ESDF-CR starts
to perform comparable to random channel selection scheme at higher data traffic.
This can be followed from the selection of higher bandwidth channels to satisfy the
bare bandwidth demand which increases with the traffic.
Figure 6.7 shows the relation between the number of blocked packets in the system
with the total number of communication segments transmitted per 100 s. The size of
Fig. 6.7 Number of packets x 10

4
2
blocked versus total number best effor data through ESDF−CR
of communication segments 1.8 background data through ESDF−CR
best effort data through random channel sel.
initiated 1.6 background data through random channel sel.
1.4
1.2
0.8
0.6
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
4
Total no. of communication segments initiated x 10
voice, video, best effort and background traffic packets are all set to 0.05 Mb with
20 primary channels. As the number of communication segments increases the total
number of blocked best effort and background traffic packets also increases. From
the figure it can be seen that a major portion (best effort and video traffic) of the
traffic is blocked in the case of ESDF-CR which can be followed from the selection
of low energy consuming channels (Fig. 6.6).
6.4 Discussion
Energy management is a critical issue in any portable system which is primarily

powered by a battery with limited endurance between recharges. The failure of inter-
mediate nodes in a network may result in route failure including critical loss of QoS.
Energy management in a cognitive radio system can be done using three steps: trans-
mission power control, spectrum allocation, and media access scheme. Transmission
power control involves limiting the transmitter output power so as to decrease the
power consumption along with the decrease in SNR/SINR. Energy consumption in a
MAC scheme is managed with the help of etiquettes designed by the access scheme.
Spectrum allocation based energy management involves selection of spectrum which
consumes lower power for transmission by selecting channels with small bandwidth
at lower frequency bands.
An energy aware spectrum decision scheme is discussed which allows cognitive
radio network to enter into low power mode whenever the power of the system falls
below a certain threshold. The scheme utilizes low power channels in low power
mode to achieve the goal of increased network longevity. Techniques for low power
channel selection along with admission control is discussed to realize the function
of energy aware spectrum decision scheme. The traffic is divided into 4 classes to
provide different access levels to different applications. The scheme selects a channel
satisfying the assured grade of QoS of the top two priority traffic while admission
control is applied on last two priority traffic. Performance analysis of the discussed
scheme is done against random channel selection scheme and the result shows that
the discussed scheme increases the longevity of the network while compensating for
large number of blocked packets for lower priority applications. Once the spectrum
selection is done, the channel IDs are passed to the media access scheme for their
utilization.
References 165
References
1. Mishra, V., Tong Lau, C., Chan, S., & Mathew, J. (2013). Energy aware spectrum decision
framework for cognitive radio network: A spectrum decision framework for cognitive radio
network with energy awareness. Journal of Low Power Electronics, 9(3), 313–321.
2. Mishra, V., Tong, L. C., Chan, S., & Kumar, A. (2012). Energy aware spectrum decision frame-
work for cognitive radio networks. In, International Symposium on Electronic System Design
(ISED) (pp. 309–313).
systems. In, IEEE 21st International Symposium on Personal Indoor and Mobile Radio Com-
munications (PIMRC) (pp. 1831–1835).
Chapter 7
Case Study: Self-coexistence
in CRN Framework
The free use of wireless frequency spectrum by cognitive radio network is anticipated
to increase its application in a multitude of wireless systems. The increase in use of
such a system will result in a problem termed as tragedy of commons as discussed
in Sect. 1.2.3. Since most of the cognitive radio systems will follow the same set of
standards (and hence similar operating frequency spectrum), they will interfere if
operated in vicinity of each other. The result will be a decrease in performance of the
system throughput along with the degradation of QoS of individual cognitive user
in the system. Mitigation of interference from similar homogeneous cognitive radio
network is termed as self-coexistence [1].
As discussed in Sect. 1.8, the mitigation of mutual interference from other homo-
geneous cognitive radio network is a 3 step process involving: resource relocation,
transmission power control, and resource sharing. Since the discussed QoS provi-
sioning framework only considers layer 2 technology of the protocol stack, trans-
mission power control for self-coexistence is beyond the scope of this book. Here,
the self-coexistence scheme for cognitive radio network are discussed from the per-
spective of resource relocation and resource sharing.
7.1 Self-coexistence Procedure
Coexistence of multiple interfering homogeneous CR networks is realized with the

help of Self-Coexistence Engine (SCE) of the discussed QoS provisioning frame-
work. The work of self-coexistence engine can be broadly classified as: detection and
recovery [2]. The SCE interacts with functional modules including success/failure
report unit and MPU to detect interference among CR networks. While SCE interacts
with CSMU, QCU and MPU to realize self-coexistence with QoS provisioning as
shown highlighted in Fig. 7.1. The procedure for self-coexistence among interfering
CR networks is discussed in following subsections.
168 7 Case Study: Self-coexistence …

stack
Application
Application QoS
Engine
Routing Module
Energy Management
Control
Self
Coexistence
Engine

Database domain
MAC instance
MAC Protocol Unit
MAC instance
Layer 2
and above
Layer 1
Success/Failure
report
Sensing report
Fig. 7.1 Modules responsible for self-coexistence in cognitive radio networks
7.1.1 Detection of Interfering Network
When a user in a CR network receives a packet associated with another CR network

(by checking network ID), it may conclude that it is in interference range of another
CR network. A similar conclusion can be drawn if a user in a CR network receives
corrupted packets for consecutive time periods. Once the interference is detected,
the interfered user in the network requests for quiet period of multiple superframe
duration on the channel in which it received the interference. During requested quiet
period, users in the network attempt to listen to packets from interfering CR net-
work/networks so as to deduce the network ID/IDs and their respective beacon start
period. Once the beacon start period is known, an interfered CR network will perform
inter-network communication to perform self-coexistence.
7.1 Self-coexistence Procedure 169
7.1.2 Recovery Process
As discussed in Sect. 1.8, an interfered user first attempts to relocate the operating
spectrum band so as to mitigate the interference. If relocation of resource is not
possible, then transmission power control (TPC) techniques are employed [1, 3]. If
TPC techniques are unsuccessful, then the resource sharing based self-coexistence
technique is applied. The resource (frequency spectrum) is shared among multi-
ple interfering CR networks by having mutually exclusive access superframe of the
channel. This is achieved by allocating each network with a non-overlapping super-
frames which are mutually exclusive. The recovery process via resource relocation
and resource sharing are discussed in subsequent sections. Recovery process requires
exchange of control messages among interfered and interfering networks along with
exchange of control messages between interfered and interfering users. Therefore
exchange of control packets can be classified as inter-network control messages and
intra-network control messages.
7.2 Resource Relocation Based Self-coexistence
As discussed in Sect. 1.8, resource in the context of cognitive radio refers to a fre-
quency band or a combination of time and frequency band. When multiple homoge-
neous cognitive radio networks operate in the vicinity, they may operate on same or
similar set of resources leading to interference. Therefore the goal of self-coexistence
from resource relocation perspective is to relocate the resources in such a way that
the interfered cognitive radio networks operate on disjoint set of resources. Com-
plexity wise, this method is more efficient and is less complex than other methods
of self-coexistence.
For the purpose of QoS provisioning, the number of resources allocated to different
networks may vary. This implies that in a resource constrained environment, a cogni-
tive radio network with heavy bandwidth demand may consume all resources leading
to resource starvation for cognitive radio network with small bandwidth demand [4].
Therefore, the system should be able to proportionally allocate resources to allow
fairness while mitigating the problem of resource starvation. Resource allocation
problems have been widely studied from different perspective including game the-
ory, graph theory, heuristic algorithm, multi-objective optimization problem to name
a few [5].
The problem of resource allocation can be directly mapped to the problem of multi-
coloring of graph and hence in this scheme, multi-coloring of graph is considered
to solve the problem of self-coexistence among multiple cognitive radio networks
from the viewpoint of resource relocation [6]. Both, resource allocation problem and
multi-coloring problem can be mapped to their corresponding optimization problems
with a given objective and applied constraints. The constraints of the resource allo-
cation problem directly maps to the constraints of the multi-coloring problem. Even
with the availability of such mapping, no formal relation has been shown between
the objectives of these optimization problems. A relationship between the objective
of these optimization problems will allow deeper understanding about the design
and analysis required to develop appropriate graph coloring heuristics. Such rela-
tionships are introduced in the subsequent sections. A multi-coloring heuristic for
self-coexistence among multiple cognitive radio networks with QoS provisioning
is also discussed. Design principles obtained from the development of relationship
between the optimization problems of resource allocation and multi-coloring is uti-
lized to develop the multi-coloring heuristic for self-coexistence.
7.2.1 System Model
The system comprises of multiple cognitive radio networks in ad-hoc mode. The total
number of different cognitive radio network is M, each comprising of Ni cognitive
users where i ∈ {1, . . . , M}. The number of channels utilized in the system is C,
each assumed to be of equal bandwidth B. The transmission range on each channel
may vary, and hence a given pair of users may interfere on some channels while
perfectly coexist on other set of channels. The set of interfered users are modeled
through graph perspective as an undirected graph represented by G = (V, E, L). V
is the set of vertices representing the interfered users, and L (L = {Lv }) is the set of
list of colors (Lv ) representing channels available at a user v ∈ V . The set of edges
representing interference between a user pair u, v ∈ V on set of channels l is given as
E = {(u, v, ), {l}}, where l ∈ {1, . . . , C}. For instance, ((x, y), {1, 3}) ∈ E implies
that user x and y interfere on channels 3 and 4, while the group ((x, y), φ) ∈ E
implies that user x and y do not interfere on any channel and hence coexist with
each other. An edge is present between two users if they interfere on at least one
channel; therefore any edge marked with group ((x, y), φ) ∈ E between vertices x
and y can be assumed to be non-existent. Neighbor of a user v is the set of groups
of user w and channels {l} ⊆ {1, . . . , C} over which user v and w interfere. This
is given by N(v) = {(w, {l|((v, w), {l}) ∈ E}), v, w ∈ V, {l} ⊆ {1, . . . , C}}. The
expression implies that the neighbor of a user v is the set of groups of user w and
channels {l}, where channels {l} satisfy the condition ((v, w), {l}) ∈ E. This work
will use the terms channel and color interchangeably, similarly vertex and user are
used interchangeably.
For instance, a graph theoretic perspective of a system of interfering cognitive
radio based ad-hoc network is shown in Fig. 7.2. The system instance comprises of
6 interfering users from 3 different cognitive radio based ad-hoc networks operating
over 4 channels {1, 2, 3, 4}. Users v1 and v4 are from network 1, users v2 and v6
are from network 2, while users v3 and v5 are from network 3. The interference
between these users is shown via edges connecting these vertices and are marked
with the colors over which the connecting users interfere. Therefore the system
comprises of edges as E = {e1 , e2 , e3 , e4 , e5 , e6 , e7 }, while e1 = ((v1 , v2 ), {2}),
e2 = ((v1 , v3 ), {3, 4}), e3 = ((v2 , v3 ), {1, . . . , 4}), e4 = ((v2 , v4 ), {1}), e5 =
7.2 Resource Relocation Based Self-coexistence 171
{2,3,4} {1,2,3,4}
v2 v6
((v2,v3),{1,…,4})
{1,2,3,4}
{2,3,4}
v1 v4
{1,2,3,4}
{1,2,3,4} v3 v5
Fig. 7.2 Graph theoretic perspective of system of interfering cognitive radio based ad-hoc networks.
The colors (channels) available at each user is shown along its corresponding vertex. The interference
between users is represented via edges between vertices marked with the channels (colors) over
which they interfere
((v3 , v4 ), {3}), e6 = ((v4 , v6 ), {1, . . . , 4}) and e7 = ((v4 , v5 ), {1, . . . , 4}). The neigh-
bor of user v3 is given as (v1 , {3, 4}), (v2 , {1, . . . , 4}), (v4 , {3}), (v5 , φ) and (v6 , φ).
7.2.2 Graph Coloring
In the previous section, mapping of the system of multiple cognitive radio network
onto an undirected interference graph was discussed. In Fig. 7.2, an instance of a
system is shown with the help of a graph, with available colors (channels) shown
along the vertices. A multi-coloring of a graph G refers to the allocation of colors
to the vertices following a certain constraint of rules. Therefore the multi-coloring
of a graph is given be set S = {Sv }, where Sv is the list of colors allocated to vertex
v ∈ V . By definition of traditional multi-coloring problem, S is a valid coloring of
G iff Su ∩ Sv = {l}, ∀((u, v), {l}) ∈ E, where u, v ∈ V , {l} ⊆ {1, . . . , C}, Su ∈ Lu
and Sv ∈ Lv [5]. Figure 7.3 shows two possible methods of multi-coloring of graph.
The colors (channels) allocated are marked with ‘*’ sign and are shown along the
vertices.
Apart from the constraint discussed above for valid coloring, additional constraints
can be imposed to achieve a certain goal. For a QoS provisioning network, additional
parameter indicating the minimum number of channels required can be introduced.
This is done so as to indicate the bandwidth demand of individual user in the system. A
new class of problem is discussed, termed as edge restricted multi-coloring (ERMC)
problem which includes an additional parameter of minimum number of channels
required with the traditional multi-coloring problem. The additional parameter is
given by M = {mv }, where mv is the minimum number of colors that need to be
allocated to vertex v ∈ V . Therefore the ERMC problem can be formulated as:
Fig. 7.3 Colors allocated

via different allocation
strategies to the graph shown
in Fig. 7.2. The colors
allocated are shown via a star
sign (*) along the vertices.
Different allocation
strategies result in different
utilization of channels
Definition 7.2 Given a graph G = (V, E, L, M), find a valid coloring of the graph
with least number of different colors used, such that each vertex v ∈ V in the graph
is colored with at least mv colors.
7.2.3 Optimization Problem Equivalent of ERMC
Let N = |V | be the total number of vertices and C be the maximum number of

different colors that are available in the system. Let the edges be represented by an
N × N × C matrix E = {ei,j,k }, where ei,j,k = {0, 1}. A value of ei,j,k = 1 implies that
an edge exists between vertices i and j and vice-versa. Therefore a value of ei,j,k = 0
implies that vertices i and j may use the same color k. The list of available colors
at each vertex is given by N × C matrix L = {li,k }. A value of li,k = 1 implies that
color k is available at vertex i and vice-versa. The additional parameter representing
minimum number of colors required at each vertex i is represented by a N × 1 vector
M = {mi }. A value of mi = δ implies that vertex i needs to be colored with at least
δ number of colors. A valid coloring of the graph is represented by N × C matrix
S = si,k . A value of si,k = 1 implies that color k is allocated to vertex i and vice-versa.
The equivalent optimization problem of the edge restricted multi-coloring (ERMC)

problem can be formulated as:

C
1, ∃i[si,k = 1]
minimize vk , where vk = (7.1)
k=1
0, otherwise
subject to si,k ≤ li,k ,

C
si,k ≥ mi ,
k=1
si,k .sj,k .ei,j,k = 0,
si,k = {0, 1},
for all i, j = 1, . . . , N, k = 1, . . . , C. Thus ERMC problem can be formulated as

a non-linear integer optimization problem. This form of optimization problem is
NP-hard in nature.
For channel allocation, the aim of a maximum throughput utility function is to
allocate channels such that the throughput of the overall system is maximized. The
optimization problem of this form of channel allocation is given as:

N
C
maximize si,k (7.2)
i=1 k=1

C
si,k ≥ mi ,
k=1
si,k = {0, 1},
for all i, j = 1, . . . , N, k = 1, . . . , C.
Both optimization problems discussed in Eqs. 7.1 and 7.2 have equivalent con-
straints imposed except for different objectives. The equivalence of constraints allow
solving the problem of resource allocation from graph theoretic perspective, except
for the objective of the optimization problem.
7.2.4 Relationship Between Optimization Problems
Work done on resource allocation from the perspective of graph coloring exploit the
fact that a solution to Eq. 7.1 will eventually satisfy all the constraints of Eq. 7.2. But it
does not imply that the solution obtained from Eq. 7.1 is optimal for Eq. 7.2. This can
be illustrated with the help of Fig. 7.3a, b where different coloring techniques have
been applied with the additional parameter of minimum number of colors required
by vertices as m1 = 3, m2 = 2, m3 = 2, m4 = 1, m5 = 1 and m6 = 1. Both
resource allocation scheme results in same chromatic number1 of 4, but have different
throughput utility of 12 and 14 respectively. Therefore, a solution obtained from
Eq. 7.1 implies minimum number of different colors used in the graph. Both Fig. 7.3a,
b are in accordance with the Eq. 7.1 and hence are optimal solutions for it. But since
they have different throughput utilities, only one of the figures will be optimal solution
for Eq. 7.2, which in this case is Fig. 7.3b (since it has the highest throughput utility
value).
Thus it can concluded that a solution of Eq. 7.1 may not be optimal for Eq. 7.2. It
can be shown that a formal relation can be introduced between Eqs. 7.1 and 7.2 by
adding the following constraint of maximum re-utilization of colors to Eq. 7.1.

N
maximize si,k , ∀k ∈ K (7.3)
i=1
Let the optimal solution of ERMC problem in Eq. 7.1 be K coloring (i.e. the chromatic
number obtained is K). Now the optimization problem of maximum throughput
utility function for channel allocation within K channels (colors) of the graph can be
formulated as:

N
K
maximize si,k (7.4)
i=1 k=1

C
si,k ≥ mi ,
k=1
si,k = {0, 1},
for all i, j = 1, . . . , N, k = 1, . . . , C.
Let the optimal solution of optimization problem in Eq. 7.1 along with the addi-
tional new constraint of Eq. 7.3 be S. Let an algorithm A be used to achieve this
optimal solution. It can be proved by contradiction, that S is also an optimal solution
to Eq. 7.2. Let S
be an optimal solution for Eq. 7.4 with K
coloring of its equivalent
graph. Therefore
N K K
N

si,k < si,k (7.5)

i=1 k=1 i=1 k=1
1 Chromatic number refers to the minimum number of different colors required to color all the
vertices in the graph following all the constraints imposed on the graph.
Two cases arise from this:

(i) When K
< K:This contradicts with the assumption that S is a solution for the
optimization problem of Eq. 7.1. Since if there already exists a coloring which
satisfies all constraints of Eq. 7.4 with K
< K, it would have been considered
by Algorithm A.
(ii) When K
= K : From this we have Ni=1 Kk=1 si,k < Ni=1 Kk=1 si,k for a fixed
N N
value of K, =⇒ s
i=1 i,k < s
i=1 i,k which contradicts with the assumption
in Eq. 7.3
Therefore solution S is optimal for Eq. 7.4. From this a more generalized optimization
function of maximum throughput utility for channel allocation can be obtained.
Suppose the coloring procedure is repeated for K1 + K2 + · · · Kn = C on disjoint
sub-graphs G 1 , G 2 , . . . G n of G for a finite n with K1 , K2 ,…Kn all disjoint to each
other. We obtain the following optimization formulation of maximum throughput
utility function:

N
C
maximize si,k (7.6)
i=1 k=1

C
si,k ≥ mi ,
k=1
si,k = {0, 1},
Thus it can be concluded that an approximation algorithm which solves the ERMC
problem with the intention of maximum re-utilization of all colors will subsequently
approximately solve the optimization problem of maximum throughput utility func-
tion of channel allocation. This result is utilized in the next section for allocating
channels based on their re-utilization.
7.2.5 Self Coexistence Scheme with QoS Provisioning
It can be safely assumed that nodes in the network support multiple priorities (cate-
gories) of data packet. Each node discovers its neighbor nodes and the channels on
which it interferes with them. Here QoS is provisioned in terms of bandwidth and
therefore a QoS metric is utilized for the purpose of calculation of bandwidth demand
of each node. QoS metric is assumed to be the weighted average of last 10 frames in
queue. Weighted average allows giving high weights to packets that were recently
queued. It also allows prioritizing the bandwidth based on the packet category. The
QoS metric (QM) of a node v in open spectrum system is calculated as:
9 T
j=0 Wi .Nij .(10 − j)
i=1
QMv = (7.7)
T 9
i=1 W i . j=0 (10 − j)
Here Nij represents the number of queued packet of type i in the last jth frame. T
denotes the number of priorities (access categories) supported by the system. Wi are
the weights assigned to packets of type i, where i ∈ {1,…,T }.
For the purpose of fairness in channel allocation, a relative share of the demand
for each node is calculated. A relative share indicates the bandwidth demand of a
node in comparison to its neighbor interfering nodes. Since channel resources are re-
utilized among all nodes of the interference graph, except among neighbor vertices;
and neighbors are defined relative to the channel on which it causes interference, we
calculate a resource share RS of a node v for channel i as:
QMv
RSv,i = , i ∈ {1, .., C}, v ∈ V (7.8)
QMv + j∈N
(v) QMj
where QMv is the QoS metric of node v and QMj is the QoS metric of neighbor nodes
of v from unique network for channel i. Here N
(v) is the set of neighbor node from
unique network on channel i. Therefore neighbor node consists of only one node
from a given network. For example in Fig. 7.2, node v3 has neighbors as (v1 , {3, 4}),
(v2 , {1, . . . , 4}) and (v4 , {3}). For resource share calculation of node v3 on channel
3, only nodes v2 and v4 are taken into consideration (since node v4 and node v1 are
from the same network). This is done to avoid allocating more share to a network
with higher number of interfered nodes. If two or more nodes from the same network,
interfere on same channel, then the node which has the highest degree for all colors
is considered for calculation. If two or more node satisfies this condition, then node
having the least number of channels available is taken into account. If nodes have
same number of channels available, then a node is randomly selected.
All nodes in the interference graph calculate a degree vector (DV), which is an
element-wise summation of all channel availability lists (CALs) received from neigh-
bor nodes including the CAL of node under consideration. Degree vector indicates
the re-utilization of channels among the neighbors of a node. Each node v ∈ V main-
tains a contention vector (CV) of size C, with each index consisting of a random value
drawn from range
DVi
CVi = 0, , ∀i ∈ {1, . . . , C} (7.9)
RSi ∗ πi
where πi is the steady state probability of channel i ∈ {1, . . . , C} to be free on node

v (discussed in Sect. 3.4.1). The design of such contention value allows selection
of channels with high availability and less usage (hence maximizes re-utilization)
by nodes having high value of relative share; while randomness in selection of con-
tention value allows mitigation of contention collision and resource deprivation. This
contention vector is exchanged among all neighbors of the interference graph. The
vertex then allocates all those channels for which the indices in contention vector
has received least value.
A distributed algorithm for the implementation of the above discussed self coex-
istence scheme is shown in Algorithm 7.1. The given heuristic executes over each
node v ∈ V and approximately solve the ERMC problem with the probability of
maximum re-utilization of colors with fairness. Calculation of QoS metric (line 1)
allows QoS prioritization among nodes. The absoluteness in greed for prioritization
is then removed by calculation of relative share RS (line 3). Degree vector (line 4)
is calculated with the help of channel availability list (CAL). CAL(j) identifies the
channel availability list of node j, whereas CALi (j) implies the availability of chan-
nel i at node j. Contention vector is calculated and shared (line 5-7) so that vertices
(nodes) win or lose colors (channels) on non-interfering basis. CVi is the value of
ith index of contention vector of node v, whereas CV (w) is the contention vector of
node w. Winning colors are then allocated to the vertex (line 8) and removed from
channel availability list (line 9) of node v and its neighbors. The algorithm terminates
on a node when all available channels on that node are allocated (lines 10-12). The
algorithm has a polynomial bound convergence time for a node, given that the node
or at least one of its neighbor node has a non-zero QoS metric.2
Algorithm 7.1: Distributed algorithm for self-coexistence executing over a

node v
1 Initialization : For a node v ∈ V , Calculate and share QoS metric from Eq. 7.7;
2 while true do
3 Calculate RS from Eq. 7.8;
N(v)
4 Calculate DVi = j=1 CALi (j), ∀i ∈ {1, . . . , C}
5 CVi = random[0, RSDV i
i ∗πi
], ∀i ∈ {1, . . . , C};
6 Transmit(CV (v));
7 Receive(CV (w)), w ∈ {N(v)};
8 colors = Select(mini (CV (v), CV (w))), ∀i ∈ {1, . . . , C};
9 CAL(w) = CAL(w)\{colors}, ∀w ∈ {v, N(v)};
10 if CALv = φ then
11 break-while;
12 end
13 end
2 Proof is provided in Appendix D.

The performance of the discussed scheme is evaluated against traditional distributed

list multi-coloring scheme [7] and utility graph coloring scheme [7] with frame level
simulation in Matlab and ideal sensing scheme. For a fair comparison, the utility
graph coloring scheme discussed in [7] is implemented in distributed way for ad-hoc
network. The analysis is done in terms of convergence time, fairness in allocation
and bandwidth deprivation of each node. Convergence time refers to the number
of successive superframes required before all channels are allocated in the system.
Bandwidth deprivation refers to average deprivation of bandwidth of each node and is
calculated as the difference between bandwidth demand and bandwidth allocated to
a node. Fairness refers to the unbiased allocation of resources (channels) to different
nodes based on their proportional need (relative share) in the system. Jain’s fairness
index is utilized over bandwidth allocation ratio (ratio of bandwidth allocated to
bandwidth demand of each node) [8]. Jain’s fairness index for N nodes with xi share
of resource allocated to each node i is calculated as:
N
( i=1 xi )
2
J(x1 , x2 , . . . , xN ) = N 2 (7.10)
N ∗ i=1 xi
The simulation generates random interference complete graph for each run with
uniform random multi-colored edges (interference). If an edge does not receive any
coloring (i.e. no interference on any channel), then it is removed from the graph. To
simplify simulation, all nodes in the graph are assumed to be from different networks.
for all channels i ∈ {1, . . . , C} and for all vertices v ∈ V . Each channel utilized in
the system is modeled as a two-state Markov chain as discussed in Sect. 3.4.1, and
the parameters α and β are chosen as uniform random values in the range [0.1, 0.9]
for all channels i ∈ {1, . . . , C} and for all vertices v ∈ V . Each node generates
4 types of Poisson modeled traffic: voice, video, best-effort and background; with
percentage distribution of mean arrival as 4, 10, 60 and 26 % respectively. The size
of voice, video, best-effort and background data packets used are 0.05, 0.5, 0.05 and
0.05 Mb respectively. Half of the nodes have mean data traffic as 4, 10, 60 and 26
packets per frame for voice, video, best-effort and background data packets; while
the other half of the nodes have 16, 40, 240 and 104 packets per frame for voice,
video, best-effort and background data packets. The bandwidth of each channel is
assumed to be of 1 Mbps. The size of superframe used in the simulation is 100 ms
and each data point in the graph is the average of 1000 runs.
Figure 7.4a, b shows convergence time (in terms of numbers of consecutive super-
frames) and fairness index (Jain’s Index) with respect to total number of nodes used in
the system. Total number of channels used is 100. The discussed coloring technique
allows faster convergence with linear relation to that of number of nodes used in the
system. This is attributed to the contention based design of resource allocation. On the
other hand utility graph coloring scheme is derived from traditional graph coloring
scheme and hence requires more convergence time than traditional coloring scheme
500
Convergence time
400
300
Discussed coloring scheme
200
Traditional coloring scheme
100 Utility graph coloring
0
2 4 6 8 10 12 14 16 18 20
No. of nodes
0.9
J Index
0.8
0.7
Utility graph coloring
2 4 6 8 10 12 14 16 18 20
No. of nodes
Fig. 7.4 Convergence time and fairness index with respect to number of nodes
in ad-hoc environment. The discussed coloring scheme also provides fair propor-
tional allocation (high value of J Index) to nodes in the system. This is attributed to
the calculation and utilization of relative share (RS) in resource allocation (Eq. 7.8).
Figure 7.5 shows the average bandwidth deprived per node for a system with 8
nodes with respect to the number of channels used in the system. The number of
channels used in the system is varied from 50 to 230. With the increase in number
of channels, the decrease in deprived bandwidth can be attributed to the increase
in number of channels allocated to a node in the system. The discussed coloring
scheme allows proportional allocation and therefore has less deprived bandwidth in
comparison to traditional allocation and utility graph coloring scheme.
7.2.7 Infrastructure Mode Operation
The resource relocation based self-coexistence technique discussed in the previous

section utilized the system model of multiple networks in ad-hoc mode. The discussed
scheme can be extended to support self-coexistence among multiple CR networks
operating in infrastructure mode. To do so, each interfered cognitive access point
80

70
Bandwidth deprived (Mbps) 60 Utility graph coloring
50
40
30
20
10
50 100 150 200 250
No. of channels
Fig. 7.5 Average bandwidth deprivation of each node in the system
is represented by a vertex while the interference between these cognitive access

points is represented by an edge. Each vertex in the generated interference graph
calculates its resource share by measuring and exchanging the QoS metric with
its neighbors. Based on the resource share, the multi-coloring scheme discussed is
utilized to allocate disjoint set of colors (channels) to each vertex in the interference
graph.
7.3 Resource Sharing Based Self-coexistence
As discussed in Sect. 1.8, resource sharing based self-coexistence schemes can be uti-
lized for the purpose of mitigation of interference between cognitive radio networks
operating in the vicinity of each other. This method of self-coexistence is applied
when resource relocation and transmission power control based self-coexistence are
not successful to mitigate the interference. The complexity of self-coexistence based
on resource sharing technique is higher than that in other techniques. Two methods
of resource sharing have been discussed: frequency division multiple access and time
division multiple access.
Since, frequency division multiple access based resource sharing is done at the
Layer 1 of the protocol stack, it is beyond the scope of this book. In time division mul-
tiple access, a resource access (in this case duration of channel access) is divided into
small time-durations. These time-durations are allocated in a disjointed way to all
the interfering users such that each user has its own exclusive non-overlapping super-
frame. An access to non-overlapping superframe allows disjointed time-duration for
channel access and hence mitigation of interference among multiple cognitive radio
7.3 Resource Sharing Based Self-coexistence 181
networks operating in the vicinity of each other [9]. The bandwidth allocated to each
network is controlled by the size of superframe allocated to each CR network. In
the subsequent sections, we discuss a QoS provisioning channel sharing based self-
coexistence protocol for multi-channel CR based networks. The discussed scheme
allows multiple homogeneous CR networks to coexist while minimizing interference
and also guaranteeing a certain grade of QoS to the applications running on different
nodes in networks.
7.3.1 System Model
The system model of multiple cognitive radio networks in ad-hoc network config-
uration is utilized. The open spectrum system comprises of M secondary ad-hoc
networks, each comprises of Ni CR nodes, ∀i ∈ M. It is assumed that neighbor
nodes follow common spectral opportunities. The system operates in an hetero-
geneous channel environment comprising of C channels, each of bandwidth Bi ,
∀i ∈ {1, . . . , C}. The system utilizes the MAC scheme discussed in Chap. 5 and the
spectrum decision scheme discussed in Sect. 4.4 of Chap. 4. If the system operates
over other set of MAC scheme and spectrum decision scheme, users in the system
should have the capability to communicate the load distribution cross-layer in proto-
col stack. This allows allocation of appropriate bandwidth to different users in case
a self-coexistence is required.
Different QoS metrics can be used depending upon the scheme used by the CR
network. Here, a QoS metric is assumed to be the weighted average of packets
sent/received in last 10 frames. The weighted average gives higher priority to trans-
mission that happened recently. This is similar to the QoS metric defined in Sect. 7.2.5
and can be calculated as:
9 T
j=0 i=1 Wi .Nij .(10 − j)
QMv = (7.11)
T 9
i=1 W i . j=0 (10 − j)
Here Nij represents the number of queued packet of type i in the last jth frame. T
denotes the number of priorities (access categories) supported by the system. Wi are
the weights assigned to packets of type i, where i ∈ {1,…,T }.
7.3.2 Design and Working
Since the discussed method of resource sharing is based on the sharing of time
duration of channel access among different secondary users, the system is assumed
to be divided into fixed length time frames. Any generic media access scheme with
the capability of division of channel access into time-durations can be modified to
Superframe i-1 Superframe i Superframe i+1
Superframe
Beacon
Beacon
Q Data transfer period Q
P P
Fig. 7.6 Superframe structure of a medium access scheme
incorporate the discussed method of self-coexistence. Figure 7.6 shows a superframe

of a generic medium access scheme comprising a beacon period and a data transfer
period. During beacon period, all users in the network exchange beacon control
messages and synchronize the whole network. During the data transfer period (DTP),
users in the network perform the data transmission. Apart from these two durations,
a cognitive radio network employs quiet period (QP) for the purpose of spectrum
sensing. This period can be made as a part of beacon or it can be separately defined
during data transfer period. Each network in the system is associated with a network
ID (NID), which is embedded in all transmissions done by users in the network as a
part of beacon.
7.3.2.1 Recovery Process
Once an interference is detected and the recovery process initiates resource sharing
based self-coexistence, the interfering users starts communicating with each other
via inter-network control messages. A resynchronization is required by one of the
interfering CR network among two CR networks who are trying to coexist with each
other, so as to mitigate any access collision. The CR network which resynchronizes
is called as foreign network and a user in that network is called as foreign user. The
network which does not perform any synchronization is called as home network and
the user in it as home user.
Four inter-network control messages are employed for the purpose of self-
coexistence:
• NWADV: Network advertisement message comprises of information about the
network including number of channels, number of users, information about oper-
ating channels, coexisting network information, beacon start period, number of
users on the current channel, quiet period distribution etc. The information about
the QoS metric is also encoded in the packet so as to enable interfering CR net-
works to calculate the size of exclusive superframe duration for the purpose of
QoS provisioning.
• NWACK: Network acknowledgement message serves as an acknowledgement to

the NWADV messages. This message ensures connectivity between coexisting CR
networks.
• SYN: Synchronization message is sent from a home user to a foreign user to
indicate that the foreign network has to resynchronize with the home network.
• SYNACK: Synchronization acknowledgement is sent by a foreign user to the home
network to indicate that the foreign network has resynchronized with the home
network and is capable of coexisting with the home network.
Figure 7.7 shows a self-coexistence session via the help of inter-network control mes-
sages. For resource sharing based self-coexistence, when a user detects interference,
it waits for a random duration of time and sends NWADV message to the interfering
user/network. This random duration allows mitigation of collision of NWADV mes-
sages which may be initiated by the other CR network. Once the NWADV message
is transmitted, the user waits for his acknowledgement (NWACK). Based on the
exchange of NWADV, a decision is made about which networks will act as the home
network and the foreign network. Once the decision is made, each interfered network
calculates its superframe size, and this superframe size is passed by the home user to
the foreign user along with a request to resynchronize its network. Once the foreign
network is synchronized with the home network, the foreign user sends an acknowl-
edgement (SYNACK). The home network verifies the synchronization through the
help of beacon start period and superframe boundaries.
Fig. 7.7 Exchange of Home NW Foreign NW

inter-network control packets
Interference
for self-coexistence detected
Random duration NWAD
V
K
NWAC
V
NWAD
NWAC Decision:
K
Decision: Foreign NW
Home NW
Calculate
Calculate
Superframe size
Superframe size
SYN
Resynchronize
network
K
SYNAC
Check
Syncrhonization
time
7.3.2.2 Home/Foreign Network Decision
For a coexistence procedure to operate, the system has to decide which CR networks
will act as the home network and which as the foreign network. During the exchange
of NWADV messages, if the message indicates that it is already coexisting with
some other CR network, then the network with whom a new network interferes is
selected as the home network. If none of the network already coexists with any other
networks, then the network which has largest number of nodes is chosen as the home
network. This is done to decrease the convergence time required by foreign network
for resynchronization.
7.3.2.3 Superframe Size Calculation
An interfered user in the network calculates the size of the superframe by utilizing
the QoS_Metric field of NWADV message. For two CR networks which are trying
to coexist with each other, a home user calculates the size of superframe as:

QMH
Size = ∗ (SF − QP) + QP (7.12)
QMH + QMF
QMH is the QoS metric of the home user while QMF is the QoS metric of the
foreign user. QP is the size of the quiet period and SF is the original superframe
size before the coexistence took place. Similarly, a foreign user also calculates the
size of coexistence superframe, except the size of quiet period is removed from the
calculation. Here, the system utilizes only one quiet period for whole time frame to
increase the throughput of the system. A foreign user learns about the distribution of
the quiet period of home user from NWADV message and utilizes it as its own quiet
period for incumbent sensing.
7.3.2.4 Virtual Synchronization
Once the superframe size calculations are done, the home node sends a SYN message
to the foreign node indicating the boundary values of its own superframe size and
also indicating that the foreign node can now resynchronize its own network in such
a way that it is virtually in sync with the home network. If the calculations done
by the foreign node agrees with the values sent by the home node, the foreign node
proceeds with the resynchronization of the beacon start period using intra-network
control packets. Once the foreign network is resynchronized, the foreign node sends
back a SYNACK message to the home node indicating that it is virtually synchronized
with the home network. To check the virtual synchronization, the home node checks
for foreign network beacon to start at the end of the home network superframe.
Similar check is done by the foreign node to ensure that there are no collisions in
frame durations. The exchange of inter-network control packets between two CR
Fig. 7.8 Coexistence Superframe

Home
procedure Network
Beacon
Beacon
P P
Superframe
Foreign
Beacon
Beacon
Network
P P
Virtual synchronization
Home Beacon Beacon

Network
Data transfer Reserved by foreign
Q Q
period network
P P
Foreign
Beacon
Network Reserved by home
Data transfer
network
period
Superframe Superframe
home network foreign network
network is shown in Fig. 7.7. A virtual synchronization process is shown in Fig. 7.8
with two homogeneous CR networks before coexistence procedure (Fig. 7.8a) and
after coexistence procedure (Fig. 7.8b).
7.3.3 Multiple CR Network Coexistence
Whenever nodes of multiple CR networks want to coexist with each other, they co-
exist sequentially one by one. Waiting for random duration of time before transmitting
NWADV message helps in avoiding collisions between multiple NWADVs. If a
node overhears a co-existence initiation session going on between two other CR
networks, it waits for SYNACK message and then initiates its own co-existence
session. Whenever a new node wants to join, all previously joined nodes have to
perform the calculation of superframe size. An example calculation of superframe
for jth foreign network can be done as following:

QMFj
Size = ∗ (SF − QP) (7.13)
QMH + M i=1 QMFi
Here M is the number of CR networks which want to coexist with each other. All
new joining networks will resynchronize their beacons to meet the boundaries of
previously joined foreign networks. The coexistence is done between only interfering
networks. For example if a system has nodes from three CR networks N W 1, N W 2
and N W 3 in which N W 1 interferes with N W 2 while N W 2 interferes with N W 3,
then co-existence is done between N W 1 − N W 2 and N W 2 − N W 3. The node of CR
network N W 2 propagates the information that it already coexist with some other CR
network and would ask the joining network to act as foreign network. The joining
node would join considering the boundary value of N W 2. Here N W 2 acts as a home
network for communication of inter-network control messages with the new joining
network.
The performance of the discussed resource sharing based self-coexistence technique

is evaluated in terms of real time connection failure with respect to total number of
communication segments used and total number of channels used in the system. The
discussed resource sharing scheme is evaluated using discrete event simulation in
Matlab which simulates the network up to frame level. All C channels are modeled
as two-state Markov chain as discussed in Sect. 3.4.1, and the parameters α and β for
channel availability are assumed to be uniformly distributed over the range [0, 1].
The bandwidth of each channel is modeled to be linear with the number of channels
used in the system, i.e. the minimum bandwidth of a channel is given a value 1 Mbps
and the maximum bandwidth of the channel is C Mbps. Therefore the bandwidth is
distributed linearly from channel 1 to channel C as 1 to C Mbps. The system utilizes
the QoS provisioning medium access scheme discussed in Chap. 5 with four access
categories. Therefore the system has 4 types of packets: voice, video, best effort
and background data. To simplify the analysis, it can be assume that two mutually
interfering homogeneous CR networks N W 1 and N W 2 are trying to coexist with each
other on all operating channels via resource sharing approach. All channels have a pair
of communicating nodes from each network. The traffic from these nodes consists of a
mixture of voice, video, best-effort and background data with percentage distribution
as 4, 10, 60 and 26 % respectively [10]. The size of voice, video, best-effort and
background packets are 0.05, 0.2, 0.05 and 0.05 Mb respectively. For performance
analysis in terms of QoS, it is assumed that network N W 2 has 3 times more traffic
intensity than that of network N W 1. The total simulation time is 100 s, whereas each
point in the graph is the average run of 100 tests. The terms- connections, packets
and segments are used interchangeably.
Figure 7.9 shows the number of voice connections failed with respect to the total
number of voice segments initiated. The total number of channels used in the system
is 10 whereas the total number of voice segments initiated is varied from 25 to 500
per second. The number of voice connections failure increases with the increase
in total number of voice segments due to the bandwidth demand exceeding the
Fig. 7.9 Number of voice 100
connections failed with 90

increase in total number of NW1 with QoS
80
communication segments NW2 with QoS
No. of voice connections failed

NW1 without QoS
70
NW2 without QoS
60
50
40
30
20
10
0
0 50 100 150 200 250 300 350 400 450 500
Total no. of voice segments initiated
available bandwidth. Networks which coexist without QoS provisioning have very
small number of voice connection failures and large number of voice connection
failures for N W 1 and N W 2 respectively. Therefore the system appears to be biased
towards the network having small traffic intensity (N W 1). The discussed resource
sharing scheme with QoS provisioning distributes the bandwidth according to the
traffic intensity, therefore the disparity between the number of voice connection failed
for both coexisting network is reduced.
Figure 7.10 shows the total number of voice connection failed including N W 1
and N W 2 with respect to total number of voice segments initiated. The total number
of channels used in the system is 10 whereas the total number of voice segments
initiated is varied from 25 to 500 per second. As the total number of voice segments
increases, the discussed approach reduces the number of voice connection failures
for the combined system of multiple networks. As evident from Figs. 7.9 and 7.10,
Fig. 7.10 Total number of 100
voice connections failed with 90

System with QoS self−coexistence
increase in total number of System without QoS self−coexistence
80
Total no. of voice connections failed
communication segments
70
60
50
40
30
20
10
0
0 50 100 150 200 250 300 350 400 450 500
Total no. of voice segments initiated
the discussed approach not only imparts bandwidth distribution according to load
but also decreases the total number of voice connections failure for the combined
system.
Figure 7.11 shows the number of best effort connections failed with respect to the
total number of best effort segments initiated. The total number of channels used in
the system is 10 whereas the total number of best effort segments initiated is varied
from 375 to 7500 per second. The number of best effort connections failure increases
with the increase in total number of best effort segments due to the bandwidth demand
exceeding the available bandwidth. N W 2 with QoS and without QoS shows equal
performance, while N W 1 without QoS shows better performance than N W 1 with
QoS. This reduced performance in lower priority packets of QoS based networks
is due to the increased performance of higher priority packets in QoS networks as
shown in Figs. 7.9 and 7.10.
Figure 7.12 shows the number of background connections failed with respect to
the total number of background segments initiated. The total number of channels
Fig. 7.11 Number of best 500
effort connections failed with 450 NW1 with QoS

400
No. of best effort connections failed
communication segments NW1 without QoS

NW2 without QoS
350
300
250
200
150
100
50
0
0 1000 2000 3000 4000 5000 6000 7000 8000
Total no. of best effort segments initiated
voice connections failed with 400 NW1 with QoS

No. of background connections failed
communication segments 350 NW1 without QoS

NW2 without QoS
300
250
200
150
100
50
0
0 500 1000 1500 2000 2500 3000 3500
Total no. of background segments initiated
used in the system is 10 whereas the total number of background segments initiated
is varied from 163 to 3250 per second. The number of background connections
failure increases with the increase in total number of background segments due
to the bandwidth demand exceeding the available bandwidth. Networks N W 1 and
N W 2 without QoS performs better than networks N W 1 and N W 2 with QoS. This
reduced performance in lower priority packets of QoS based networks is due to
the increased performance of higher priority packets in QoS networks as shown in
Figs. 7.9 and 7.10.
In Fig. 7.13, an analysis of the total number of voice connection failed with the
number of channels used in the system is shown. The traffic consists of 15, 37, 225
and 95 communication segments of voice, video, best-effort and background data
which gives the percentage distribution as 4, 10, 60 and 26 % for voice, video, best-
effort and background data respectively. The number of channels is varied from 2 to
22. From the graph, it can be observed that the number of channels have no effect on
the number of voice connection failures in the networks. Also the discussed approach
continues to show a decrease in disparity between the connection failures of home
and foreign network.
In Fig. 7.14, an analysis of the total number of voice connections failed with the
total number of channel used in the system is done. The traffic consists of 15, 37,
225 and 95 communication segments of voice, video, best-effort and background
data which gives the percentage distribution as 4, 10, 60 and 26 % for voice, video,
best-effort and background data respectively. The number of channels is varied from
2 to 22. From the graph, it can be observed that the discussed approach reduces the
combined number of voice connection failure in the system.
Fig. 7.13 Number of voice 50

connections failed with 45
different number of channels
40 NW1 with QoS
used
No. of voice connections failed
NW2 with QoS

35 NW1 without QoS
30 NW2 without QoS
25
20
15
10
0
2 4 6 8 10 12 14 16 18 20 22
No. of channels

voice connections failed with 50
different number of channels
Total no. of voice connections failed

48
used
46
44 System with QoS self co−existence

System without QoS self co−existence
42
40
38
36
34
32
2 4 6 8 10 12 14 16 18 20 22
No. of channels
7.4 Discussion
The open spectrum system promises free usage of spectrum on an opportunistic

basis. But this free usage will eventually suffer from a problem termed as tragedy of
commons, which will increase the interference among cognitive radio based wireless
system and thus restricting their growth. To mitigate such scenario, self-coexistence
techniques have been introduced which mitigates interference among similar homo-
geneous cognitive radio networks. Three different forms of self-coexistence tech-
niques exists: resource relocation, transmission power control and resource sharing.
Resource relocation and resource sharing based interference mitigation techniques
are applied at Layer 2 of the protocol stack, while transmission power control based
interference mitigation is applied at Layer 1 of the protocol stack.
A multi-coloring based resource relocation policy for self-coexistence in cognitive
radio networks is discussed which allows for the formation of a relationship between
edge restricted multi-coloring (ERMC) scheme and maximum throughput utility
function of channel allocation. A distributed scheme for solving ERMC problem is
discussed with the intention of maximum re-utilization. The multi-coloring based
resource relocation policy allows fair allocation of channels while serving the QoS
demands of individual nodes. It has been shown that the discussed scheme allows
fairness with proportional allocation and converges much faster than utility graph
coloring and traditional coloring scheme.
A frame sharing based self-coexistence procedure for cognitive radio networks
is also discussed from the perspective of resource sharing. The discussed scheme
allows generic medium access scheme to incorporate the feature of self-coexistence
using little modification. The scheme allows two or more homogeneous CR net-
works to virtually synchronize and share media with each other on a non-interfering
basis. The sharing is done using modified superframe size which depends upon the
QoS requirement of the interfered node of a CR network. The frame sharing based
approach for resource sharing allows fair calculation of frame size while serving the
7.4 Discussion 191
QoS demands of individual users. It has been shown that the discussed scheme not
only impart bandwidth distribution according to load but also decreases real time
connections failure of combined system.
References
1. Lee, D. H., Jeon, W. S., Jeong, D. G., & Ko, G. (2010). Self-coexistence techniques for cognitive
radio lans/pans. In IEEE 21st International Symposium on Personal Indoor and Mobile Radio
Communications (PIMRC), (pp. 1516–1520).
2. Vishram, M., Tong, L. C., & Syin, C. (2015). A channel allocation based self-coexistence
scheme for homogeneous ad-hoc networks. IEEE Wireless Communications Letters, 4, 545–
548.
3. Hoven, N., & Sahai, A. (2005). Power scaling for cognitive radio. In International Conference
on Wireless Networks, Communications and Mobile Computing, 1, (pp. 250–255).
4. Pattara-Aukom, W., Banerjee, S., & Krishnamurthy, P. (2002). Starvation prevention and quality
of service in wireless lans. In The 5th International Symposium on Wireless Personal Multi-
media Communications, 3, (pp.1078–1082).
5. Han, R., & Liu, K. (2008). Resource allocation for wireless networks: basics, techniques, and
applications. Cambridge: Cambridge University Press.
6. Vishram, M., Tong, L. C., & Syin, C. (2014). List multi-coloring based fair channel allocation
policy for self coexistence in cognitive radio networks with qos provisioning. In Region 10
Symposium, 2014 IEEE, (pp.99–104).
7. Sengupta, S., Brahma, S., Chatterjee, M., & Shankar, N. S. (2013). Self-coexistence among
interference-aware IEEE 802.22 networks with enhanced air-interface. Pervasive and Mobile
Computing, 9(4), 454–471.
8. Jain, R., Durresi, A., & Babic, G. (1999). Throughput fairness index: An explanation, Technical
Report, Department of CIS, The Ohio State University.
9. Vishram, M., Tong, L. C., & Syin, C. (2013). Qos provisioning self co-existence protocol for
cognitive radio ad-hoc networks. In 19th IEEE International Conference on Networks (ICON),
(pp.1–6).
10. Mishra, V., Tong, L. C., & Syin, C. (2012). Qos based spectrum decision framework for
cognitive radio networks. In 18th IEEE International Conference on Networks (ICON), (pp.18–
23).
Appendix A
Proof of Lemma 4.2
Algorithm 4.1 based on the greedy approach of selecting the longest consecutive free
time-slots provides an optimal solution to the minimal channel switching problem.
We provide a proof by contradiction by considering a vector V, which is obtained
using the discussed greedy approach mentioned in the Algorithm 4.1. Let there exist
a vector U which is not obtained by the greedy approach and has
η−1 η−1

Switching(Uk ) ≤ Switching(Vk ) (A.1)
k=0 k=0
where η = (24 ∗ 60/x) − 1 and x is the sampling interval in minutes.

Also, si ze(U) = si ze(V) = total time-units
= (24 ∗ 60/x) − 1 = constant (A.2)
Let m and n represents the total number of channel switching required when the
system utilizes vector U and V for channel selection respectively. Let αi and γ j
represent the vector indices blocks which have consecutively same channels in vector
U and V respectively; the values i and j identify a particular block in vector U and
V respectively. The values of i and j range from 1 to m + 1 and n + 1 respectively.
We assume that αi = 1 and γ j = 1 for all possible values of i and j. We also define
time blocks ti and θ j as the total time-units to which the system remains in blocks
αi and γ j respectively. For example, if channel selection of a vector U is given by [1
1 3 3 2 2 2 3 3], then the total number of switching m = 3; total time-units in blocks
t1 = 2, t2 = 2, t3 = 3 and t4 = 2. Now from Eq. A.2

m+1
n+1
si ze(U) = si ze(V) =⇒ αi ti = γjθj (A.3)
i=1 j=1
From Eqs. A.1 and A.3 =⇒ m < n

Signals and Communication Technology, DOI 10.1007/978-3-319-45860-1
194 Appendix A: Proof of Lemma 4.2
Now, if m < n and αi = γ j = 1, ∀i ∈ {1, . . . , m + 1}, j ∈ {1, . . . , n + 1}, the

given equality in Eq. A.3 =⇒ ∃i∃ j : ti > θ j .
This contradicts with the assumption that the vector V is obtained using greedy
approach. If there exists any time block ti greater than θ j , then that would have been
taken by our greedy approach.
Appendix B
Proof of Lemma 4.3
To prove that High Throughput Minimal Switching (HTMS) problem is in NP, we

show that a solution to this problem can be verified in polynomial time. Let an
instance of solution be vector V. To check whether it satisfies the constraints (total
number of switching < δ and bandwidth > ω) or not, we need to check:
η−1
(i) k=0 Switching(Vk ) < δ
η
(ii) k=0 BW (Vk ) > ω, where BW (i) returns the bandwidth of channel i
(iii) MVk ,k = 1, ∀k ∈ η , where M is channel availability matrix
Verifying
(i) Takes a running time of O(η)
(ii) Takes a running time of O(η + 1)
(iii) Takes a running time of O(η + 1)
Thus an instance of the solution can be verified in polynomial time.

Appendix C
Proof of Lemma 4.4
To prove that High Throughput Minimal Switching (HTMS) problem is NP-hard, we

will reduce a known NP-hard problem into HTMS problem. We consider Linearly
Constrained Binary Quadratic Programming (LCBQP) which is a known NP-hard
problem [1, 2]. Let the LCBQP be of the form:

n
n
f (x) = x t Qx = qi, j xi x j xi ∈ {0, 1}∀i (C.1)
i=1 j=1
subject to a x < b.
We modify this LCBQP problem by considering vectors consisting of binary
values instead of variables xi , x j etc., which makes it easier to reduce it into our
problem. Let the solution to HTMS problem be represented by a vector VS. We
convert this solution vector into solution matrix MS, which consists of rows and
columns representing the channels ids and time-units respectively. A value of one
inside the matrix means that the corresponding row is selected for that column i.e. a
particular channel is selected for that particular time-unit.
i.e.
1, if V j = i
MSi, j = ∀i ∈ C , ∀ j ∈ η (C.2)
0, otherwise
where i, j represent rows and columns for matrix MS; similarly, for vector V, j
represents its index. Let X j be the vector representing the jth column of the solution
matrix MS. Now we formulate the objective function and various constraints. We
define the objective function to be minimum switching, satisfying a given bound on
channel bandwidth.
The minimum switching is given by maximizing the following quadratic function.
Switching = X0T .X1 + X1T .X2 + X2T .X3 + X3T .X4 +

(C.3)
X4T .X5 + . . . Xη − 1T .Xη

198 Appendix C: Proof of Lemma 4.4
We assume each channel has bandwidth Bc , where c is the channel index. We assume
a vector BV of size C, consisting of bandwidths of all channel and is given as:
BV = [B1 B2 B3 B4 B5 B6 B7 . . . BC ]
Therefore, overall bandwidth for 24 h is given as,
BW = BV.X0 + BV.X1 + BV.X2 + BV.X3 + BV.X4 +

(C.4)
BV.X5 + . . . BV.Xη
We define bandwidth BW to be greater than some constant ω. Now we describe the

validity constraints of the system. For a solution vector VS to be valid, it needs to
satisfy three constraints:
(i) The vectors X0 , X1 , X2 , X3 , X4 , X5 . . . etc. are valid only if these satisfy the
availability of channels in channel availability matrix M. We define M j as the
vector of column j of matrix M. Therefore from this constraint we have
M0T .X0 + M1T .X1 + M2T .X2 + M3T .X3 + M4T .X4 +
(C.5)
. . . MηT .Xη = η + 1
(ii) There could be only one index having a value of 1 in each solution vector X.
Therefore from this constraint we have,
1ηT .X j = 1, ∀ j ∈ η (C.6)
where 1η is a row vector of 1 of size η

(iii) Also, Xi ∈ {0, 1}, ∀i, is a constraint that needs to be satisfied for binary value
of vector elements.
Therefore from Eqs. C.3, C.4, C.5 and C.6 we get,
Maximize:
X0T .X1 + X1T .X2 + X2T .X3 + X3T .X4 + . . . Xη − 1T .Xη
Subject to:
BV.X0 + BV.X1 + BV.X2 + BV.X3 + . . . BV.Xη > ω,
and M0T .X0 + M1T .X1 + M2T .X2 + . . . MηT .Xη = η + 1,
and 1ηT .X j = 1, ∀ j ∈ η and Xi ∈ {0, 1}, ∀i
From Eq. C.1 we have

n
n
f (x) = x t Qx = qi, j xi x j , xi ∈ {0, 1}, ∀i
i=1 j=1
Subject to a x < b

1, if i + 1 = j
Assume qi, j = ∀i∀ j
0, otherwise
Appendix C: Proof of Lemma 4.4 199
Even after this assumption the problem remains in the class of NP-hard [2],
n n
=⇒ f (x) = i=1 j=1 qi, j x i x j
=⇒ f (x) = x0 x1 + x1 x2 + x2 x3 + x3 x4 + x4 x5 + . . . xn−1 xn
subject to a x < b
This form of LCBQP problem represents the HTMS problem we have defined with
variables replaced by vectors. Thus the LCBQP problem is reduced to the HTMS
problem and hence lies in the domain of NP-hard problems.
References
1. Garey, M. R., & Johnson, D. S. (1990). Computers and Intractability; A Guide to the Theory of
NP-Completeness. New York, NY, USA: W. H. Freeman & Co.
2. Merz, P., Freisleben, B. (1999) Genetic algorithms for binary quadratic programming. In
GECCO-1999: Proceedings of the Genetic and Evolutionary Computation Conference, pp.
417–424. Morgan Kauffman.
Appendix D
Proof of Polynomial Bound Convergence
of Algorithm 7.1
Lemma D.1 Algorithm 7.1 converges in polynomial time for a node v, given that it
has non-zero QoS metric (Q M). In other words, the proposed algorithm has polyno-
mial bound convergence time for any node which has sent/received any data traffic
in the last 10 frames.
Proof Consider an undirected graph G = (V, E, L), with V , E and L as its vertices,
edges between vertices and list of colors (containing C colors) available at each vertex
respectively. For a node v under consideration with non-zero weights W in Eq. 7.7;
any data traffic sent/received in last 10 frames will result in non-zero QoS metric
(Q M) for node v. And therefore the resource share calculated from it will be also
non-zero. Algorithm 7.1 converges iff list C AL becomes empty. Therefore to prove
Algorithm 7.1 converges it is necessary and sufficient to prove that all colors from
list C AL v are either allocated to node v or its neighbor N (v) in polynomial time.
For worst case scenario, the contention vector C V will never win the contention for
selection of colors (lines 5–8 of Algorithm 7.1). But for every non-winning contention
of node v, some other neighbor node w (w ∈ N (v)) may win the contention. If no
nodes in N (v) win the contention, then neighbor nodes of w i.e. N(N(v)) and so
on, will eventually win the contention. And therefore, every winning contention in
neighbor nodes will remove the colors from C AL. Since the algorithm is distributed
in nature and all nodes executes concurrently, the maximum iterations by which all
colors from list C AL are removed is determined by the diameter of the graph. In
the worst case scenario, the maximum diameter is attained by a Path Graph with the
diameter of N − 1 [1]. Therefore the worst case convergence time of the algorithm
is N − 1 iterations, and hence it has polynomial bounded convergence time.
Lemma D.2 Algorithm 7.1 converges in polynomial time for a node v having zero
QoS metric, given that for each color l available in C AL, node v has at least one
neighbor w with non-zero QoS metric and l ∈ C AL. In other words, if a node v has
zero value for QoS metric (Q M), it can still converge in polynomial time given that
it shares common spectral opportunity with at least one neighbor with non-zero QoS
metric for every color.

202 Appendix D: Proof of Polynomial Bound Convergence of Algorithm 7.1
Proof Consider an undirected graph G = (V, E, L), with V , E and L as its vertices,
edges between vertices and list of colors (containing C colors) available at each vertex
respectively. For a node v under consideration with non-zero weights W in Eq. 7.7;
if no data traffic is sent/received in last 10 frames, it will result in zero QoS metric
(Q M) for node v. But since, node v shares common spectral opportunity with at
least one neighbor with non-zero QoS metric for every color, the resource share
(Eq. 7.8) is zero instead of non-defined. Thus the calculation of contention vector
(line 5 of Algorithm 7.1) is a defined value. Since resource share is zero, the value
of contention vector generally selected is large and hence has a low probability of
winning the contention. In worst case scenario, the contention vector C V will never
win the contention for selection of colors (lines 5–8 of Algorithm 7.1). This has been
reduced to the proof of Lemma D.1, where it has been shown that the Algorithm 7.1
converges for worst case scenario where contention vector never wins the contention.
Therefore this algorithm has polynomial bound convergence time.
Reference
1. Caccetta, L., & Smyth, W. (1992). Graphs of maximum diameter. Discrete Mathematics, 102(2),
121–141.

Coronavirus (COVID-19) #LivUniCOVID

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Coronavirus (COVID-19) #LivUniCOVID

Uploaded by

Copyright:

Available Formats

Signals and Communication Technology

QoS and Energy

QoS and Energy

ISSN 1860-4862 ISSN 1860-4870 (electronic)

© Springer International Publishing Switzerland 2017

Printed on acid-free paper

This Springer imprint is published by Springer Nature

provides different ways through which energy can be efﬁciently managed in

Chapter 3 Media Access in CRN

The structure of this book is given as shown in ﬁgure above. It is recommended

Singapore, Singapore Vishram Mishra

2 Cognitive Radio Network- A Review . . . . . . . . . . . . . . . . . . . . . . .... 39

4 Case Study: Spectrum Management in CRN Framework . . . . . . . . . 111

6 Case Study: Energy Management in CRN Framework . . . . . . . . . . . 155

Recently, the rapid proliferation of portable wireless communication devices sup-

Fig. 1.1 Opportunistic utilization of spectral holes

model, where licensed (primary/incumbents) and unlicensed (secondary/CR) users

1.1 Why This Book?

mentalization of different units allows easy abstraction, debugging and maintenance

1.2 Spectrum Regulation

International Telecommunication Union (ITU) [17] is a specialized agency of the

1.2.1 Licensed Spectrum

1.2.2 Unlicensed Spectrum

30GHz 3GHz 300MHz 30MHz 3MHz 300kHz 3kHz

Aero Rad 74.6 - 4.438 -

4.8 - 400.05 - Radio Sat 75.4 - 4.85 -

5.25 - Aero Rad 406.1 - Fixed

RadioLocate 410 - Radio Astro 5.45 - 415 -

Aero Rad 5.73 -

5.47 - 5.95 - Fixed

5.85 - 6.765 - Aeronatical

5.875 - 7.1 - Amateur

470 - 118 - 7.35 -

494 - 8.195 - 526.5 -

137 - 9.04 - Aeronatical

7.75 - Amateur 11.4 - Aeronatical

7.9 - 790 - 146 - Fixed 130 -

8.4 - 890 - 12.1 -

8.85 - 150.05 - Radio Sat 13.36 - Aeronatical

162 - 14.99 - Mobile

10 - LM Fix 15.01 - 1800 -

Mob Sat 17.48 - Fixed

13.25 - Mob Sat Fixed

1555 - 18.168 - Amateur

14.8 - 230 - 19.68 -

15.35 - Fixed 1631.5 - Maritime 2107 -

if this type of spectrum is free-to-use, it is regulated in terms of interference policies

1.2.3 Open Spectrum

1.3 Opportunistic Spectrum Usage

Fig. 1.3 Average spectrum occupancy in Singapore

1.4 Software Defined Radio and Cognitive Radio

Conventional radio communication systems have components implemented in hard-

1.4.1 IEEE Groups Working on Spectrum Sharing

1. 802.11a: This group is responsible for designing Dynamic Frequency Selection

1.4.2 Cognition Cycle

Fig. 1.4 Cognition cycle for

Decision Radio sense

1.4.3 Cognitive Engine and Framework

A cognitive engine (CE) is formulated and implemented on various SDR boards in

Fig. 1.5 A typical cognitive

Radio Framework: Radio framework acts as a middleware in between cognitive

1.4.4 Cognitive Radio Network

1.5 Quality of Service (QoS)

1. Access Category: To provide packet priority, packets need to be classified into

Pr {|1 − pz (n)| < ε} > 1 − γ, ∀n > n